1
00:00:04,960 --> 00:00:20,000
[Music]

2
00:00:21,199 --> 00:00:24,000
it's the student

3
00:00:27,640 --> 00:00:33,800
showcase I am ridiculously excited this

4
00:00:30,279 --> 00:00:35,360
is definitely the highlight of my year

5
00:00:33,800 --> 00:00:37,640
any other speakers in the room don't

6
00:00:35,360 --> 00:00:38,879
don't don't take offense okay the way

7
00:00:37,640 --> 00:00:41,200
this works we've got a whole bunch of

8
00:00:38,879 --> 00:00:45,440
awesome speakers they're all students

9
00:00:41,200 --> 00:00:50,239
from all across Australia but also local

10
00:00:45,440 --> 00:00:56,879
um we have nine speakers presenting

11
00:00:50,239 --> 00:00:56,879
seven talks 10 minutes one fire

12
00:00:57,640 --> 00:01:01,680
possible we'll find out we're putting

13
00:00:59,760 --> 00:01:03,039
them last so that everyone else gets a

14
00:01:01,680 --> 00:01:06,280
chance to talk in case we all get

15
00:01:03,039 --> 00:01:09,159
evacuated from the building so without

16
00:01:06,280 --> 00:01:13,960
further Ado fire extinguishers maybe

17
00:01:09,159 --> 00:01:16,560
talk next year um okay this is summer uh

18
00:01:13,960 --> 00:01:19,920
she has been programming since she was

19
00:01:16,560 --> 00:01:22,320
nine her favorite emoji is the Space

20
00:01:19,920 --> 00:01:25,280
Invaders

21
00:01:22,320 --> 00:01:26,799
Emoji um and something cool that you

22
00:01:25,280 --> 00:01:28,560
found out about computers recently is

23
00:01:26,799 --> 00:01:31,799
that you learned that computer viruses

24
00:01:28,560 --> 00:01:34,000
weren't originally harmful uh the

25
00:01:31,799 --> 00:01:37,240
creeper in 1971 just displayed I'm the

26
00:01:34,000 --> 00:01:40,680
creeper Catch Me If You Can cool and

27
00:01:37,240 --> 00:01:40,680
this is your talk summer thank you very

28
00:01:42,560 --> 00:01:49,799
much okay hello everyone so this is my

29
00:01:46,240 --> 00:01:52,000
ghost to got Chief first of all hi my

30
00:01:49,799 --> 00:01:54,640
name is Sumer I'm a student at Bergman

31
00:01:52,000 --> 00:01:57,079
Anin school and I'm going to go to UTS

32
00:01:54,640 --> 00:01:59,000
for cyber security next year um I've

33
00:01:57,079 --> 00:02:02,119
done work experience with Microsoft and

34
00:01:59,000 --> 00:02:04,479
examplify I've also am a part of protest

35
00:02:02,119 --> 00:02:06,920
Stars this season and last season um

36
00:02:04,479 --> 00:02:09,720
which is with Blackbird I went to ncss

37
00:02:06,920 --> 00:02:13,480
summer school in Sydney for this year

38
00:02:09,720 --> 00:02:15,800
and I recently won the wick women in ICT

39
00:02:13,480 --> 00:02:17,120
Kate M student encouragement award for

40
00:02:15,800 --> 00:02:20,120
this

41
00:02:17,120 --> 00:02:20,120
year

42
00:02:20,840 --> 00:02:26,680
wo so as an introduction the porogi is

43
00:02:24,879 --> 00:02:29,680
basically this device which was created

44
00:02:26,680 --> 00:02:31,959
by evil socket in 2019 he's a developer

45
00:02:29,680 --> 00:02:34,440
and and security researcher and was

46
00:02:31,959 --> 00:02:37,160
inspired by the popular 1990s toy the

47
00:02:34,440 --> 00:02:39,720
tomag GOI obviously first before you say

48
00:02:37,160 --> 00:02:41,360
anything porn is not a bad word it's

49
00:02:39,720 --> 00:02:43,120
what happens when you try to spell own

50
00:02:41,360 --> 00:02:44,780
but can't and it's a video game thing

51
00:02:43,120 --> 00:02:49,400
from ages ago

52
00:02:44,780 --> 00:02:52,040
[Laughter]

53
00:02:49,400 --> 00:02:55,480
apparently so it's basically a portable

54
00:02:52,040 --> 00:02:57,640
device it's AI powered um and uses used

55
00:02:55,480 --> 00:03:00,920
for Wi-Fi penetration testing it's

56
00:02:57,640 --> 00:03:03,120
operated by capturing WPA and WPA2

57
00:03:00,920 --> 00:03:04,840
handshakes from nearby networks helping

58
00:03:03,120 --> 00:03:07,440
ethical hackers identify potential

59
00:03:04,840 --> 00:03:09,040
security flaws the device stands out

60
00:03:07,440 --> 00:03:11,120
because of its userfriendly design

61
00:03:09,040 --> 00:03:13,920
portability and its interactive

62
00:03:11,120 --> 00:03:15,959
personality it not only educates users

63
00:03:13,920 --> 00:03:17,680
about cyber security but also makes

64
00:03:15,959 --> 00:03:20,280
learning processes engaging through its

65
00:03:17,680 --> 00:03:21,720
tomagotchi inspired interface it's a

66
00:03:20,280 --> 00:03:24,080
compelling tool because of what it can

67
00:03:21,720 --> 00:03:26,319
do and its interactive nature especially

68
00:03:24,080 --> 00:03:27,599
for students and cyber security

69
00:03:26,319 --> 00:03:29,439
enthusiasts

70
00:03:27,599 --> 00:03:31,120
standpoint from an educational

71
00:03:29,439 --> 00:03:33,360
standpoint point the point ofi is highly

72
00:03:31,120 --> 00:03:35,319
appropriate it simplifies complex

73
00:03:33,360 --> 00:03:37,360
security tasks making them accessible to

74
00:03:35,319 --> 00:03:39,760
beginners however its Effectiveness

75
00:03:37,360 --> 00:03:42,080
depends on user responsibility basically

76
00:03:39,760 --> 00:03:42,080
don't be

77
00:03:42,560 --> 00:03:48,519
illegal so the way it works is it uses

78
00:03:45,159 --> 00:03:50,680
AI integration um that learns optimal

79
00:03:48,519 --> 00:03:52,879
strategies from capturing WPA hand

80
00:03:50,680 --> 00:03:55,200
checks it's portable it runs on

81
00:03:52,879 --> 00:03:59,079
lightweight Hardware such as the raspber

82
00:03:55,200 --> 00:04:00,640
pi Z Wave share V4 like I used or the V3

83
00:03:59,079 --> 00:04:01,840
and a battery

84
00:04:00,640 --> 00:04:04,959
it's also interactive through its

85
00:04:01,840 --> 00:04:08,159
displays and shows emotion based

86
00:04:04,959 --> 00:04:10,879
performance here is the interface for my

87
00:04:08,159 --> 00:04:13,360
porag GOI um it's very cute as you can

88
00:04:10,879 --> 00:04:15,639
see uh the porag GOI itself is down the

89
00:04:13,360 --> 00:04:18,120
bottom and the other ones are the web

90
00:04:15,639 --> 00:04:19,560
UI but before we get more into it I just

91
00:04:18,120 --> 00:04:22,400
want to talk about the

92
00:04:19,560 --> 00:04:24,160
importance so cyber secur knowledge is

93
00:04:22,400 --> 00:04:26,360
critical in today's world for

94
00:04:24,160 --> 00:04:28,160
individuals it protects personal data

95
00:04:26,360 --> 00:04:30,360
for organizations it prevents costly

96
00:04:28,160 --> 00:04:32,639
breaches and for society it helps create

97
00:04:30,360 --> 00:04:34,759
a safer digital environment tools like

98
00:04:32,639 --> 00:04:36,880
the pornog GOI make cybercity Concepts

99
00:04:34,759 --> 00:04:38,880
accessible and relatable they Empower

100
00:04:36,880 --> 00:04:40,080
users to take control of their online

101
00:04:38,880 --> 00:04:41,880
security while highlighting the

102
00:04:40,080 --> 00:04:44,199
vulnerabilities we Face

103
00:04:41,880 --> 00:04:46,360
daily this is why learning about cyber

104
00:04:44,199 --> 00:04:48,039
security isn't just for professionals

105
00:04:46,360 --> 00:04:50,880
it's something everyone should care

106
00:04:48,039 --> 00:04:52,840
about the porag Gotti's focus on Wi-Fi

107
00:04:50,880 --> 00:04:55,080
security makes it an interesting entry

108
00:04:52,840 --> 00:04:57,120
point for cyber security education its

109
00:04:55,080 --> 00:04:58,759
Effectiveness depends on how well users

110
00:04:57,120 --> 00:05:00,160
understand and apply its lessons to a

111
00:04:58,759 --> 00:05:02,639
broader context

112
00:05:00,160 --> 00:05:09,960
some statistics there were

113
00:05:02,639 --> 00:05:09,960
2365 cyber attacks in 2023 with 343 mil

114
00:05:28,639 --> 00:05:34,280
338,000 an ethical hacking matters it

115
00:05:32,560 --> 00:05:36,280
identifies vulnerabilities and Wi-Fi

116
00:05:34,280 --> 00:05:38,720
networks raises awareness about

117
00:05:36,280 --> 00:05:41,199
protecting personal organizations and

118
00:05:38,720 --> 00:05:44,160
data encourages proactive defenses

119
00:05:41,199 --> 00:05:46,880
against cyber threats however Simplicity

120
00:05:44,160 --> 00:05:48,680
is the porag Go's double-edge sword

121
00:05:46,880 --> 00:05:51,000
while it lowers the barrier for entry

122
00:05:48,680 --> 00:05:52,960
into cyber security it also requires

123
00:05:51,000 --> 00:05:54,919
users to have a strong ethical

124
00:05:52,960 --> 00:05:56,840
foundation without proper guidance

125
00:05:54,919 --> 00:05:58,880
there's obviously a risk for

126
00:05:56,840 --> 00:06:00,199
misuse there's also learning through the

127
00:05:58,880 --> 00:06:02,319
poroi tool

128
00:06:00,199 --> 00:06:04,440
it offers a hands-on experience Bridges

129
00:06:02,319 --> 00:06:06,800
the gap between theoretical knowledge

130
00:06:04,440 --> 00:06:09,000
and practical application and teaches

131
00:06:06,800 --> 00:06:10,800
critical problem solving skills and of

132
00:06:09,000 --> 00:06:12,080
course we want to raise awareness about

133
00:06:10,800 --> 00:06:15,479
Wi-Fi

134
00:06:12,080 --> 00:06:17,280
security so my project ex itself aimed

135
00:06:15,479 --> 00:06:19,400
to build on the pondi's strengths by

136
00:06:17,280 --> 00:06:22,000
adding gamification features such as

137
00:06:19,400 --> 00:06:23,680
achievements and growth systems these

138
00:06:22,000 --> 00:06:25,520
additions would have increased user

139
00:06:23,680 --> 00:06:28,000
engagement and encourage continuous

140
00:06:25,520 --> 00:06:30,160
learning unfortunately though I fac

141
00:06:28,000 --> 00:06:32,000
significant challenges integrating the

142
00:06:30,160 --> 00:06:34,160
features into the pornog GOI special I

143
00:06:32,000 --> 00:06:35,560
code based and it proved much more

144
00:06:34,160 --> 00:06:37,400
complex than I

145
00:06:35,560 --> 00:06:39,039
anticipated while I couldn't fully

146
00:06:37,400 --> 00:06:40,880
Implement these features I gained

147
00:06:39,039 --> 00:06:41,960
valuable insight into debugging and

148
00:06:40,880 --> 00:06:44,319
software

149
00:06:41,960 --> 00:06:46,280
design this experience highlighted the

150
00:06:44,319 --> 00:06:47,639
importance of planning adaptability and

151
00:06:46,280 --> 00:06:50,000
learning from

152
00:06:47,639 --> 00:06:52,199
failure while my project didn't meet all

153
00:06:50,000 --> 00:06:54,360
its objectives it reinforced the porag

154
00:06:52,199 --> 00:06:55,680
Go's Effectiveness as a learning tool

155
00:06:54,360 --> 00:06:56,560
and I hope to eventually get these

156
00:06:55,680 --> 00:06:58,639
features

157
00:06:56,560 --> 00:07:00,440
working so going back to the actual

158
00:06:58,639 --> 00:07:02,919
pornog GOI itself

159
00:07:00,440 --> 00:07:07,039
again it's very cute but here's a

160
00:07:02,919 --> 00:07:09,120
working video so when the porag GOI says

161
00:07:07,039 --> 00:07:11,039
let's be friends it basically means it

162
00:07:09,120 --> 00:07:14,440
has detected another pornog GOI device

163
00:07:11,039 --> 00:07:16,960
nearby initiating a social interaction

164
00:07:14,440 --> 00:07:19,240
this is basically based of the porogi uh

165
00:07:16,960 --> 00:07:21,919
the tomag GOI like personality

166
00:07:19,240 --> 00:07:24,280
system the channels in the context of a

167
00:07:21,919 --> 00:07:26,199
porogi refers to the Wi-Fi

168
00:07:24,280 --> 00:07:29,199
channels sorry that audio is not

169
00:07:26,199 --> 00:07:30,840
supposed to be working anyway um it

170
00:07:29,199 --> 00:07:33,360
refers first the Wi-Fi channel is used

171
00:07:30,840 --> 00:07:34,800
by wireless networks for communication

172
00:07:33,360 --> 00:07:38,080
um the Wi-Fi channels are specific

173
00:07:34,800 --> 00:07:40,599
frequency ranges within 2.1 GHz and 5

174
00:07:38,080 --> 00:07:42,520
gahz Spectrum used by wireless devices

175
00:07:40,599 --> 00:07:46,120
to communicate the

176
00:07:42,520 --> 00:07:50,560
porogi the porogi usually stays between

177
00:07:46,120 --> 00:07:53,080
1 and 14 as those are 2.4 gahz channels

178
00:07:50,560 --> 00:07:55,039
um they also shows the number of access

179
00:07:53,080 --> 00:07:56,960
points of the top and down the bottom

180
00:07:55,039 --> 00:08:01,159
where it says pawned it is the number of

181
00:07:56,960 --> 00:08:01,159
handshakes during the current session

182
00:08:01,280 --> 00:08:04,960
so first of all I just want to say I

183
00:08:02,840 --> 00:08:07,639
struggled to get the code that I had

184
00:08:04,960 --> 00:08:10,080
made into the poroi operating

185
00:08:07,639 --> 00:08:12,319
system uh these are some of the issues

186
00:08:10,080 --> 00:08:14,159
by the way the reason I don't have the

187
00:08:12,319 --> 00:08:16,919
porag GOI up and running in front of me

188
00:08:14,159 --> 00:08:19,599
at the moment um is Du to the cord being

189
00:08:16,919 --> 00:08:20,840
very very short for the battery system

190
00:08:19,599 --> 00:08:22,440
um because my teacher believes in

191
00:08:20,840 --> 00:08:25,599
something called a budget for some

192
00:08:22,440 --> 00:08:28,680
strange reason

193
00:08:25,599 --> 00:08:33,560
boo um also the original porag GOI ran

194
00:08:28,680 --> 00:08:35,240
from wave sh uh V3 screen but I got a V4

195
00:08:33,560 --> 00:08:37,039
thinking it was the same turns out

196
00:08:35,240 --> 00:08:38,279
they're different and they're different

197
00:08:37,039 --> 00:08:41,200
enough that the libraries don't want to

198
00:08:38,279 --> 00:08:43,839
play nice anymore from this I obviously

199
00:08:41,200 --> 00:08:45,920
learned a lot I've done game development

200
00:08:43,839 --> 00:08:48,839
before and I'm used to working in Python

201
00:08:45,920 --> 00:08:51,399
however I'm not used to working in pi or

202
00:08:48,839 --> 00:08:54,040
Powershell so again it made it a bit

203
00:08:51,399 --> 00:08:56,320
hard so here's some errors as you can

204
00:08:54,040 --> 00:08:59,160
see this eror is the log page where the

205
00:08:56,320 --> 00:09:01,760
poroi would be working but not actually

206
00:08:59,160 --> 00:09:03,240
receiving handshakes um to fix this I

207
00:09:01,760 --> 00:09:05,200
basically had to reflash the whole

208
00:09:03,240 --> 00:09:08,560
system a couple hundred times when

209
00:09:05,200 --> 00:09:10,720
making this this next one was an error

210
00:09:08,560 --> 00:09:12,680
from network connection my computer had

211
00:09:10,720 --> 00:09:14,880
told me that the porag GOI was connected

212
00:09:12,680 --> 00:09:16,800
to my home network but after further

213
00:09:14,880 --> 00:09:19,079
inspection the connection was just way

214
00:09:16,800 --> 00:09:20,839
too weak for the porag GOI to run and it

215
00:09:19,079 --> 00:09:23,880
was just making cute faces not actually

216
00:09:20,839 --> 00:09:23,880
doing any handshake

217
00:09:24,600 --> 00:09:29,600
capturing and lastly we had an updating

218
00:09:27,240 --> 00:09:32,320
issue this was because the porter got

219
00:09:29,600 --> 00:09:36,720
system I was using was so it was much

220
00:09:32,320 --> 00:09:38,279
newer than the original one oh sorry um

221
00:09:36,720 --> 00:09:42,519
and the device didn't want to get the

222
00:09:38,279 --> 00:09:44,480
plugins working so it made a few weird

223
00:09:42,519 --> 00:09:46,240
errors um unfortunately I still haven't

224
00:09:44,480 --> 00:09:48,880
been able to fix this just because again

225
00:09:46,240 --> 00:09:50,959
I can't update the

226
00:09:48,880 --> 00:09:53,560
systems but what I've learned from this

227
00:09:50,959 --> 00:09:56,560
is continuous debugging and researching

228
00:09:53,560 --> 00:09:59,399
actually proved a deeper insight into um

229
00:09:56,560 --> 00:10:01,000
the poi's functionality and even failed

230
00:09:59,399 --> 00:10:02,800
projects contribute to understanding and

231
00:10:01,000 --> 00:10:05,480
skill development and I definitely

232
00:10:02,800 --> 00:10:07,279
learned a lot from this project just for

233
00:10:05,480 --> 00:10:09,200
some insight this is some of the plugins

234
00:10:07,279 --> 00:10:11,519
that I had constructed using basic

235
00:10:09,200 --> 00:10:15,480
python um and it was just a class

236
00:10:11,519 --> 00:10:18,320
structure and some defining of codes um

237
00:10:15,480 --> 00:10:18,320
this was for the

238
00:10:20,800 --> 00:10:25,680
achievements but basically to conclude

239
00:10:23,160 --> 00:10:27,040
the porogi is a small but powerful tool

240
00:10:25,680 --> 00:10:29,560
that demonstrates the importance of

241
00:10:27,040 --> 00:10:32,360
cyber security while its focus is is

242
00:10:29,560 --> 00:10:34,560
narrow it is effective and educates

243
00:10:32,360 --> 00:10:37,279
users about Wi-Fi vulnerabilities and

244
00:10:34,560 --> 00:10:39,160
ethical hacking practices the my project

245
00:10:37,279 --> 00:10:41,120
faced challenges but it underscores the

246
00:10:39,160 --> 00:10:43,600
value of persistence and the importance

247
00:10:41,120 --> 00:10:45,920
of learning through failure tools like

248
00:10:43,600 --> 00:10:47,880
the porag GOI remind us of the need for

249
00:10:45,920 --> 00:10:49,639
proactive security measures and

250
00:10:47,880 --> 00:10:51,519
continuous education in a rapidly

251
00:10:49,639 --> 00:10:53,160
evolving field and I would love to keep

252
00:10:51,519 --> 00:10:58,560
working on this and continue to show you

253
00:10:53,160 --> 00:10:58,560
guys how it updates thank you

254
00:10:59,360 --> 00:11:03,200
was

255
00:11:00,279 --> 00:11:04,920
amazing Brian stop clapping you are up

256
00:11:03,200 --> 00:11:08,000
next I guess I'm going have to kick you

257
00:11:04,920 --> 00:11:09,279
off stage that is fantastic um if you

258
00:11:08,000 --> 00:11:12,079
have

259
00:11:09,279 --> 00:11:14,600
questions keep them in your mind or if

260
00:11:12,079 --> 00:11:16,920
you're as bad as that as I am write them

261
00:11:14,600 --> 00:11:18,800
down because at the end there will

262
00:11:16,920 --> 00:11:22,600
hopefully be some time for questions for

263
00:11:18,800 --> 00:11:23,959
all of the students next up is Brian

264
00:11:22,600 --> 00:11:28,160
Brian

265
00:11:23,959 --> 00:11:31,279
is um Brian's favorite emoji is a thumbs

266
00:11:28,160 --> 00:11:33,560
up emoji nice one that I can actually do

267
00:11:31,279 --> 00:11:37,279
um Brian's been programming for one to

268
00:11:33,560 --> 00:11:40,320
two years and something that Brian has

269
00:11:37,279 --> 00:11:43,000
recently found out about computers is

270
00:11:40,320 --> 00:11:44,160
vectorization specifically you can

271
00:11:43,000 --> 00:11:47,880
actually reduce the amount of

272
00:11:44,160 --> 00:11:50,079
calculations you run by just uh by just

273
00:11:47,880 --> 00:11:54,440
performing operations on an entire array

274
00:11:50,079 --> 00:11:58,120
at a time cool Brian take it

275
00:11:54,440 --> 00:11:59,760
away all right hello everybody uh I will

276
00:11:58,120 --> 00:12:02,760
be here talking to you guys about how I

277
00:11:59,760 --> 00:12:04,399
use Python to find Fibonacci numbers so

278
00:12:02,760 --> 00:12:06,760
for those who don't know if you let F

279
00:12:04,399 --> 00:12:09,360
ofn be the n's Fibonacci number you find

280
00:12:06,760 --> 00:12:12,560
F ofn by adding together F of n minus

281
00:12:09,360 --> 00:12:15,519
one and F of n minus 2 so as a baseline

282
00:12:12,560 --> 00:12:18,240
f of0 is z and F1 and F2 are the same in

283
00:12:15,519 --> 00:12:20,279
that they are both one so the problem

284
00:12:18,240 --> 00:12:23,480
arises with this uh how do you find F

285
00:12:20,279 --> 00:12:25,440
ofn with code now for context at this

286
00:12:23,480 --> 00:12:26,600
point in my life I was in year 11 and I

287
00:12:25,440 --> 00:12:28,680
learned everything that I'm about to

288
00:12:26,600 --> 00:12:30,680
tell you about in this my algorithmics

289
00:12:28,680 --> 00:12:32,240
class and at that point I was finding

290
00:12:30,680 --> 00:12:33,639
the content pretty difficult to

291
00:12:32,240 --> 00:12:35,160
understand I was kind of getting cooked

292
00:12:33,639 --> 00:12:36,880
so hopefully I don't replicate that with

293
00:12:35,160 --> 00:12:38,480
you guys as I go into an explanation of

294
00:12:36,880 --> 00:12:41,720
the first solution which involves a

295
00:12:38,480 --> 00:12:43,720
little something called recursion now uh

296
00:12:41,720 --> 00:12:45,519
as a simple definition recursion is a

297
00:12:43,720 --> 00:12:47,959
tactic where you use simpler versions of

298
00:12:45,519 --> 00:12:49,959
our problem to CC to solve a more

299
00:12:47,959 --> 00:12:51,680
complex variants of our problem and in

300
00:12:49,959 --> 00:12:53,760
programming this normally involves a

301
00:12:51,680 --> 00:12:55,440
function calling itself until it reaches

302
00:12:53,760 --> 00:12:56,959
something called a base case the

303
00:12:55,440 --> 00:12:58,519
simplest iteration of a problem that's

304
00:12:56,959 --> 00:13:01,639
normally so simple that it's a

305
00:12:58,519 --> 00:13:03,800
no-brainer to answer so here is the

306
00:13:01,639 --> 00:13:06,160
first solution for calculating F ofn

307
00:13:03,800 --> 00:13:07,680
I've called it RF of n uh as you can see

308
00:13:06,160 --> 00:13:09,880
those first two if statements up there

309
00:13:07,680 --> 00:13:12,399
represent our base cases F0 equals 0 and

310
00:13:09,880 --> 00:13:14,079
FS1 and FS2 are both one and if we call

311
00:13:12,399 --> 00:13:17,000
anything that's not a base case that's

312
00:13:14,079 --> 00:13:20,800
cool we simply return rfn min-1 plus rfn

313
00:13:17,000 --> 00:13:24,079
minus 2 as a mirror of fnal FN minus1

314
00:13:20,800 --> 00:13:25,720
plus FN minus 2 so this solution works

315
00:13:24,079 --> 00:13:28,240
and it looks pretty elegant but it does

316
00:13:25,720 --> 00:13:31,360
have a weakness namely in its complexity

317
00:13:28,240 --> 00:13:33,040
now uh to put it simply again complexity

318
00:13:31,360 --> 00:13:34,920
complexity basically describes how well

319
00:13:33,040 --> 00:13:37,480
a program will scale in memory usage or

320
00:13:34,920 --> 00:13:39,279
run time as input sizes get bigger for

321
00:13:37,480 --> 00:13:41,600
the purposes of this presentation I will

322
00:13:39,279 --> 00:13:43,560
be analyzing the coming programs in Big

323
00:13:41,600 --> 00:13:45,320
O notation or worst case scenario where

324
00:13:43,560 --> 00:13:46,959
it has to do the most amount of work and

325
00:13:45,320 --> 00:13:50,320
how and I'll be analyzing how runtime

326
00:13:46,959 --> 00:13:52,279
scales alongside this so if we look at

327
00:13:50,320 --> 00:13:54,279
the amount of function calls RF ofn

328
00:13:52,279 --> 00:13:56,440
needs to make with linearly increasing

329
00:13:54,279 --> 00:13:58,800
ends you notice that uh we're starting

330
00:13:56,440 --> 00:13:59,800
to observe sort of an exponential is

331
00:13:58,800 --> 00:14:01,199
growth

332
00:13:59,800 --> 00:14:03,839
with uh the amount of function calls

333
00:14:01,199 --> 00:14:05,279
that I have to make which implies that

334
00:14:03,839 --> 00:14:08,199
the amount of time RF will need to

335
00:14:05,279 --> 00:14:10,440
execute will scale exponentially in size

336
00:14:08,199 --> 00:14:13,000
so to actually prove this what I did was

337
00:14:10,440 --> 00:14:15,639
I made RF of n calculate F0 all the way

338
00:14:13,000 --> 00:14:17,480
to F40 print out how much time it took

339
00:14:15,639 --> 00:14:19,279
for each function to complete SLA down

340
00:14:17,480 --> 00:14:21,720
on Google Sheets and made a little trend

341
00:14:19,279 --> 00:14:24,399
line and I produced these two graphs

342
00:14:21,720 --> 00:14:26,040
right here so to the to the right of the

343
00:14:24,399 --> 00:14:28,440
slide you'd see the output of that

344
00:14:26,040 --> 00:14:30,160
previous program over here uh and to the

345
00:14:28,440 --> 00:14:31,440
left the slide you see a little

346
00:14:30,160 --> 00:14:33,320
calculation that I did which should

347
00:14:31,440 --> 00:14:34,759
measure around how many function calls

348
00:14:33,320 --> 00:14:36,680
that RF needs to actually calculate

349
00:14:34,759 --> 00:14:38,800
given Fibonacci number now both of these

350
00:14:36,680 --> 00:14:41,720
look like exponentials which tell us

351
00:14:38,800 --> 00:14:44,600
that RF has a Time complexity of class

352
00:14:41,720 --> 00:14:45,880
O2 to the^ of n which meaning that if

353
00:14:44,600 --> 00:14:47,240
you plot it out in the graph like this

354
00:14:45,880 --> 00:14:48,920
it looks like an exponential and it will

355
00:14:47,240 --> 00:14:50,440
continue to scale as an exponential as n

356
00:14:48,920 --> 00:14:51,759
increases compared to what we can

357
00:14:50,440 --> 00:14:53,720
actually do with this problem this

358
00:14:51,759 --> 00:14:55,320
solution kind of sucks in terms of time

359
00:14:53,720 --> 00:14:57,320
complexity so I'm going to try my best

360
00:14:55,320 --> 00:14:58,839
to guide you to the logic of the second

361
00:14:57,320 --> 00:15:01,399
solution because this took me a really

362
00:14:58,839 --> 00:15:03,360
long time to understand so if you look

363
00:15:01,399 --> 00:15:04,839
at the process for calculating F of 5

364
00:15:03,360 --> 00:15:06,440
you notice that we have nine function

365
00:15:04,839 --> 00:15:09,920
calls here and you notice that we need

366
00:15:06,440 --> 00:15:12,680
to call F3 twice so if we associate

367
00:15:09,920 --> 00:15:14,279
something here with F3 uh and Define it

368
00:15:12,680 --> 00:15:16,639
the first time we actually calculate it

369
00:15:14,279 --> 00:15:18,040
we can actually substitute its usage its

370
00:15:16,639 --> 00:15:19,560
usage the next time we need it which

371
00:15:18,040 --> 00:15:21,360
takes our function calls down from 9 to

372
00:15:19,560 --> 00:15:22,839
7 which should ideally make our program

373
00:15:21,360 --> 00:15:25,480
a bit faster because it's calling less

374
00:15:22,839 --> 00:15:27,399
functions uh this is the idea behind our

375
00:15:25,480 --> 00:15:29,240
second solution which involves something

376
00:15:27,399 --> 00:15:30,759
called dynamic programming know the

377
00:15:29,240 --> 00:15:32,560
memorization part I do not have enough

378
00:15:30,759 --> 00:15:34,959
time to get into that

379
00:15:32,560 --> 00:15:36,959
so you can think of dynamic programming

380
00:15:34,959 --> 00:15:38,480
as a sort of fancy version of recursion

381
00:15:36,959 --> 00:15:40,800
where instead of just discarding the

382
00:15:38,480 --> 00:15:42,319
simpler cases you are storing them and

383
00:15:40,800 --> 00:15:45,199
using them later on so you don't have to

384
00:15:42,319 --> 00:15:47,040
recalculate them and in certain problems

385
00:15:45,199 --> 00:15:48,800
where these smaller values appear a lot

386
00:15:47,040 --> 00:15:50,920
in calculating these bigger values like

387
00:15:48,800 --> 00:15:52,600
in finding F of n this will save you a

388
00:15:50,920 --> 00:15:55,440
lot of time

389
00:15:52,600 --> 00:15:57,000
so again I just want to Simply preface

390
00:15:55,440 --> 00:15:58,480
that we're simply storing stuff that

391
00:15:57,000 --> 00:15:59,959
we've already done to save time so we

392
00:15:58,480 --> 00:16:02,839
don't have to calculate everything over

393
00:15:59,959 --> 00:16:04,680
and over again so if we let upgraded FIB

394
00:16:02,839 --> 00:16:07,560
be of a function that uses this strategy

395
00:16:04,680 --> 00:16:09,639
to return F ofn we get a function that

396
00:16:07,560 --> 00:16:11,680
looks like this so this looks a bit

397
00:16:09,639 --> 00:16:12,959
complicated to those to those who don't

398
00:16:11,680 --> 00:16:14,800
who haven't really caught on to what I

399
00:16:12,959 --> 00:16:16,040
was explaining but I'll hopefully I can

400
00:16:14,800 --> 00:16:17,720
do a good job at explaining it right

401
00:16:16,040 --> 00:16:19,959
here so here's the storing and

402
00:16:17,720 --> 00:16:21,199
remembering part we Define a record with

403
00:16:19,959 --> 00:16:22,480
all of the Fibonacci numbers that we

404
00:16:21,199 --> 00:16:24,279
currently have of course we start off at

405
00:16:22,480 --> 00:16:25,920
our base cases cuz those are obvious and

406
00:16:24,279 --> 00:16:27,959
the first thing we do when we call our

407
00:16:25,920 --> 00:16:29,680
function is check if it's already in our

408
00:16:27,959 --> 00:16:30,839
record if it's already our record well

409
00:16:29,680 --> 00:16:32,800
that's all cool we can just take it from

410
00:16:30,839 --> 00:16:34,720
our record and return it if not well

411
00:16:32,800 --> 00:16:37,639
that's also cool we can simply perform F

412
00:16:34,720 --> 00:16:39,639
of n = f of n minus1 plus F of n minus 2

413
00:16:37,639 --> 00:16:42,759
shove down our record and return it as

414
00:16:39,639 --> 00:16:45,319
required so to actually figure out its

415
00:16:42,759 --> 00:16:48,000
time complexity what I did was calculate

416
00:16:45,319 --> 00:16:50,880
was make a upgraded FIB calculate F0 all

417
00:16:48,000 --> 00:16:53,959
the way to F 100,000 in increments of

418
00:16:50,880 --> 00:16:55,839
2,500 so doing so gives me these two

419
00:16:53,959 --> 00:16:58,880
graphs uh gives me the graph to the left

420
00:16:55,839 --> 00:17:02,000
over here as you can see uh it takes up

421
00:16:58,880 --> 00:17:04,199
upgraded FIB 0.17 seconds to calculate F

422
00:17:02,000 --> 00:17:06,720
100,000 as opposed to RF taking 10

423
00:17:04,199 --> 00:17:08,760
seconds to calculate F of 40 but I

424
00:17:06,720 --> 00:17:10,280
recognize that we that some of you guys

425
00:17:08,760 --> 00:17:12,240
need a bit more of a mathematically

426
00:17:10,280 --> 00:17:13,880
rigorous way of comparing the two so

427
00:17:12,240 --> 00:17:15,280
let's look at the function calls so

428
00:17:13,880 --> 00:17:18,280
let's look at the function calls as well

429
00:17:15,280 --> 00:17:21,439
so I'm about to run a model which should

430
00:17:18,280 --> 00:17:23,520
uh pretty accurately uh visualize how

431
00:17:21,439 --> 00:17:25,520
many functions that upgraded FIB needs

432
00:17:23,520 --> 00:17:28,520
to call to to calculate a given

433
00:17:25,520 --> 00:17:31,200
Fibonacci number and as you can see here

434
00:17:28,520 --> 00:17:32,440
it mostly like a straight line uh with

435
00:17:31,200 --> 00:17:34,600
if you look at the longest chain of

436
00:17:32,440 --> 00:17:36,440
nodes here with the exception of one of

437
00:17:34,600 --> 00:17:37,679
the base cases down here the actual

438
00:17:36,440 --> 00:17:39,200
Fibonacci number that we're trying to

439
00:17:37,679 --> 00:17:41,960
calculate and the case that we've

440
00:17:39,200 --> 00:17:44,520
already calculated and saved uh most of

441
00:17:41,960 --> 00:17:46,640
these have one branch which tells us

442
00:17:44,520 --> 00:17:48,559
that it's going to scale mostly linearly

443
00:17:46,640 --> 00:17:50,280
in terms of function calls namely if

444
00:17:48,559 --> 00:17:52,360
we're not calling a base case upgraded

445
00:17:50,280 --> 00:17:54,880
FIB is going to call 2 N minus 3

446
00:17:52,360 --> 00:17:58,159
functions which if you plot that on a

447
00:17:54,880 --> 00:18:00,720
graph gives you a gives you a gives you

448
00:17:58,159 --> 00:18:02,880
two graphs that show a similar disparity

449
00:18:00,720 --> 00:18:06,280
with if you compare them with runtime as

450
00:18:02,880 --> 00:18:08,520
well uh it takes upgraded FIB less than

451
00:18:06,280 --> 00:18:12,400
slightly less than 200,000 functions to

452
00:18:08,520 --> 00:18:15,440
calculate F of 100,000 where uh ourf of

453
00:18:12,400 --> 00:18:17,520
n needs like what uh 200 million

454
00:18:15,440 --> 00:18:19,480
function calls to calculate the 40th

455
00:18:17,520 --> 00:18:23,000
Fibonacci number

456
00:18:19,480 --> 00:18:24,960
and analyzing this solution because we

457
00:18:23,000 --> 00:18:27,120
took the because we did the simple

458
00:18:24,960 --> 00:18:29,120
shortcut of just not calculating stuff

459
00:18:27,120 --> 00:18:31,240
that we've already done we have taken an

460
00:18:29,120 --> 00:18:33,440
exponentially scaling solution and we

461
00:18:31,240 --> 00:18:35,480
reduced it down to a linearly scaling

462
00:18:33,440 --> 00:18:38,600
solution to the point where this new

463
00:18:35,480 --> 00:18:42,159
solution in calculating the 100,000

464
00:18:38,600 --> 00:18:45,120
Fibonacci number is 60 times faster than

465
00:18:42,159 --> 00:18:48,640
our old solution calculating the 40th

466
00:18:45,120 --> 00:18:51,400
Fibonacci number so uh I might have

467
00:18:48,640 --> 00:18:54,760
rushed this talk a bit hope that's fine

468
00:18:51,400 --> 00:18:56,880
but uh the moral of the story is that uh

469
00:18:54,760 --> 00:18:58,440
we this this kind of improvement

470
00:18:56,880 --> 00:19:00,320
happened because we took a step back

471
00:18:58,440 --> 00:19:02,039
from because I took a step back from

472
00:19:00,320 --> 00:19:03,679
looking at the Fibonacci numbers problem

473
00:19:02,039 --> 00:19:05,679
realized that these smaller Fibonacci

474
00:19:03,679 --> 00:19:09,240
numbers appeared a lot in calculating

475
00:19:05,679 --> 00:19:11,400
bigger Fibonacci numbers and so and so

476
00:19:09,240 --> 00:19:13,600
adapted to that and so sort of responded

477
00:19:11,400 --> 00:19:15,520
to that aspect of the problem so this

478
00:19:13,600 --> 00:19:17,760
really shows me the importance of

479
00:19:15,520 --> 00:19:19,320
actually thinking things out before

480
00:19:17,760 --> 00:19:20,799
programming something and actually

481
00:19:19,320 --> 00:19:24,120
understanding what you're doing as

482
00:19:20,799 --> 00:19:27,799
opposed to just well uh monkey brain

483
00:19:24,120 --> 00:19:30,080
here's a loop let's hope it goes well uh

484
00:19:27,799 --> 00:19:32,240
I might have rush just a bit too hard

485
00:19:30,080 --> 00:19:35,679
but that's really all I have to say

486
00:19:32,240 --> 00:19:35,679
thank you for listening to my Pon

487
00:19:38,120 --> 00:19:45,200
presentation thank you so much monkey

488
00:19:41,679 --> 00:19:47,840
brain here's a loop Ain't That A vibe

489
00:19:45,200 --> 00:19:50,919
monkey brain here's a

490
00:19:47,840 --> 00:19:55,919
loop I think we've all been there and

491
00:19:50,919 --> 00:20:00,080
may well be again next up is Jun Jun has

492
00:19:55,919 --> 00:20:03,679
been programming for seven years

493
00:20:00,080 --> 00:20:07,360
and his favorite emoji is the I should

494
00:20:03,679 --> 00:20:07,360
make you all guess what this is based

495
00:20:09,200 --> 00:20:17,240
on yes melting oh dedication as I put my

496
00:20:12,559 --> 00:20:19,440
back out okay um uh and something that

497
00:20:17,240 --> 00:20:22,159
Jun recently learned about

498
00:20:19,440 --> 00:20:25,720
computers is that he should be taking

499
00:20:22,159 --> 00:20:30,080
dis backups after his Linux disc

500
00:20:25,720 --> 00:20:32,799
corrupted itself the day before in this

501
00:20:30,080 --> 00:20:36,679
form that's a lesson we all could learn

502
00:20:32,799 --> 00:20:38,720
too take it away J thank you hello so

503
00:20:36,679 --> 00:20:41,080
today I'll be presenting my project

504
00:20:38,720 --> 00:20:44,360
which is um scanning a train with motion

505
00:20:41,080 --> 00:20:46,919
tracking in a phone so um for some

506
00:20:44,360 --> 00:20:49,720
background I was reading Wikipedia once

507
00:20:46,919 --> 00:20:51,880
and I came across this Wikipedia page um

508
00:20:49,720 --> 00:20:54,520
and on this page there was this picture

509
00:20:51,880 --> 00:20:55,760
by the Wikipedia user Theo I found it

510
00:20:54,520 --> 00:20:58,240
really interesting because the

511
00:20:55,760 --> 00:21:00,320
background is really line like filled

512
00:20:58,240 --> 00:21:02,200
with line lines and it has a really

513
00:21:00,320 --> 00:21:05,520
sci-fi shading to it which I wanted to

514
00:21:02,200 --> 00:21:07,120
replicate so how would I do that so

515
00:21:05,520 --> 00:21:08,960
apparently according to the description

516
00:21:07,120 --> 00:21:12,600
this was taken using a technique called

517
00:21:08,960 --> 00:21:15,559
strip photography and it's a technique

518
00:21:12,600 --> 00:21:19,320
where you take one direction space and

519
00:21:15,559 --> 00:21:22,360
One Direction time so what how you do

520
00:21:19,320 --> 00:21:25,200
this is you take a continuous line like

521
00:21:22,360 --> 00:21:26,559
this and then you stack it together into

522
00:21:25,200 --> 00:21:29,200
a full

523
00:21:26,559 --> 00:21:31,440
picture and

524
00:21:29,200 --> 00:21:35,320
this was first done using film so how

525
00:21:31,440 --> 00:21:38,919
you do it with film is you use a slit

526
00:21:35,320 --> 00:21:41,520
um uh yeah you use a slit lens instead

527
00:21:38,919 --> 00:21:43,039
of like a normal lens and this allows

528
00:21:41,520 --> 00:21:46,320
you to

529
00:21:43,039 --> 00:21:50,279
capture single like single lines

530
00:21:46,320 --> 00:21:51,679
together and then like combine them and

531
00:21:50,279 --> 00:21:53,720
this is great and all but I don't have a

532
00:21:51,679 --> 00:21:56,480
film camera so how would I do it with

533
00:21:53,720 --> 00:21:58,799
with my phone um there's also digital

534
00:21:56,480 --> 00:22:02,520
cameras that do this but they cost like

535
00:21:58,799 --> 00:22:05,880
of dollars even like used ones

536
00:22:02,520 --> 00:22:09,559
so I looked I looked at my phone and my

537
00:22:05,880 --> 00:22:12,480
camera um cameras are very expensive my

538
00:22:09,559 --> 00:22:14,720
phone's free my phone can use USB the

539
00:22:12,480 --> 00:22:16,440
line scan camera has to use obscure

540
00:22:14,720 --> 00:22:19,600
interfaces oh by the way um these

541
00:22:16,440 --> 00:22:21,720
cameras are called line scan cameras but

542
00:22:19,600 --> 00:22:24,960
and the problem is they have very high

543
00:22:21,720 --> 00:22:28,400
frame rate and my phone only has 240

544
00:22:24,960 --> 00:22:31,120
frames per second so how do I compensate

545
00:22:28,400 --> 00:22:34,960
for for this well the sensor of my phone

546
00:22:31,120 --> 00:22:36,960
is very wide 4,000 pixels wide much

547
00:22:34,960 --> 00:22:40,880
wider than the one pixel wide line scan

548
00:22:36,960 --> 00:22:44,240
camera so what I decided on doing is

549
00:22:40,880 --> 00:22:47,240
instead of using um single Pixel slices

550
00:22:44,240 --> 00:22:49,960
I take multipixel slices in the middle

551
00:22:47,240 --> 00:22:54,559
and I stack them into a final

552
00:22:49,960 --> 00:22:57,600
picture so this animation so um each of

553
00:22:54,559 --> 00:23:00,480
these uh each of these frames they're

554
00:22:57,600 --> 00:23:02,240
frames in like a video and I can take

555
00:23:00,480 --> 00:23:04,400
slices in the middle and stack them

556
00:23:02,240 --> 00:23:07,279
together into a final picture so I tried

557
00:23:04,400 --> 00:23:08,159
this with a crude python script and this

558
00:23:07,279 --> 00:23:10,720
was the

559
00:23:08,159 --> 00:23:12,679
result it looks very similar to the

560
00:23:10,720 --> 00:23:14,720
first one but you can see there's some

561
00:23:12,679 --> 00:23:17,919
problems for example there's the

562
00:23:14,720 --> 00:23:20,080
contraction in the train in the top part

563
00:23:17,919 --> 00:23:22,320
there's the dilation in the top in the

564
00:23:20,080 --> 00:23:26,120
bottom left and then if you zoom in

565
00:23:22,320 --> 00:23:30,039
further you can see it's not pretty the

566
00:23:26,120 --> 00:23:32,960
the PTV logo is sort of um stretched out

567
00:23:30,039 --> 00:23:35,919
and you can you can maybe see the um

568
00:23:32,960 --> 00:23:39,720
lines in the middle that's because of

569
00:23:35,919 --> 00:23:42,000
that's just because of issues but um so

570
00:23:39,720 --> 00:23:43,919
what I how I ended up fixing this was

571
00:23:42,000 --> 00:23:47,039
using motion tracking but I don't know

572
00:23:43,919 --> 00:23:48,799
how to do motion tracking so I got a

573
00:23:47,039 --> 00:23:53,279
library to do it for

574
00:23:48,799 --> 00:23:55,760
me um this so this is open CV which is a

575
00:23:53,279 --> 00:23:57,559
computer vision uh computer vision

576
00:23:55,760 --> 00:24:01,279
Library it's available for Python and

577
00:23:57,559 --> 00:24:05,480
C++ but I don't know C++ so I use Python

578
00:24:01,279 --> 00:24:08,400
so it um I use the orb key Point finder

579
00:24:05,480 --> 00:24:10,200
The Brute Force keypoint match and I use

580
00:24:08,400 --> 00:24:14,559
I use it also for general image

581
00:24:10,200 --> 00:24:16,240
processing so first with key points so

582
00:24:14,559 --> 00:24:18,840
key points are points in the picture

583
00:24:16,240 --> 00:24:21,400
which are really identifiable and really

584
00:24:18,840 --> 00:24:23,360
easy to find like in multiple pictures

585
00:24:21,400 --> 00:24:26,480
so in this case it's like the rivets on

586
00:24:23,360 --> 00:24:31,279
the front of the train or the um the

587
00:24:26,480 --> 00:24:33,720
dots on the like like the name plate of

588
00:24:31,279 --> 00:24:35,600
the train but it also captures the pass

589
00:24:33,720 --> 00:24:37,080
on the station which is not good because

590
00:24:35,600 --> 00:24:40,840
I don't want the station I want the

591
00:24:37,080 --> 00:24:43,360
train so what open CV allows allows you

592
00:24:40,840 --> 00:24:47,760
to do is it allows you to key Point

593
00:24:43,360 --> 00:24:50,320
match between two frames and with this I

594
00:24:47,760 --> 00:24:54,200
can plot out the distance that key Point

595
00:24:50,320 --> 00:24:57,240
moves like between the frame and

596
00:24:54,200 --> 00:24:58,799
if on on the vertical axis if the

597
00:24:57,240 --> 00:25:00,760
vertical change in position and the

598
00:24:58,799 --> 00:25:04,240
horizontal axis is a horizontal change

599
00:25:00,760 --> 00:25:07,880
in position and you can see it's very

600
00:25:04,240 --> 00:25:10,720
noisy there's a lot of like bad data the

601
00:25:07,880 --> 00:25:12,640
red ones are filtered out by putting an

602
00:25:10,720 --> 00:25:13,799
absolute limit on the vertical axis

603
00:25:12,640 --> 00:25:17,399
because the train doesn't move

604
00:25:13,799 --> 00:25:19,760
vertically it only moves horizontally um

605
00:25:17,399 --> 00:25:22,799
the standard deviation filtering is used

606
00:25:19,760 --> 00:25:25,279
to filter out the horizontal outliers

607
00:25:22,799 --> 00:25:28,000
and what that gives me is the two points

608
00:25:25,279 --> 00:25:29,840
in the middle the blue ones but the blue

609
00:25:28,000 --> 00:25:33,240
dots aren't actually single points there

610
00:25:29,840 --> 00:25:35,399
are a lot of points and you can see

611
00:25:33,240 --> 00:25:38,039
there's two clums the one on the left is

612
00:25:35,399 --> 00:25:39,720
the train because it's moving in the X

613
00:25:38,039 --> 00:25:42,760
Direction you can see the bottom part

614
00:25:39,720 --> 00:25:45,120
shows minus 10 which isus 10 pixels move

615
00:25:42,760 --> 00:25:47,880
this Frame and then the station doesn't

616
00:25:45,120 --> 00:25:49,720
move at all but the train like the

617
00:25:47,880 --> 00:25:51,760
computer doesn't have a concept of a

618
00:25:49,720 --> 00:25:53,840
clump like we do so how do we

619
00:25:51,760 --> 00:25:55,600
distinguish between them so I use a

620
00:25:53,840 --> 00:25:58,360
technique called kernel density

621
00:25:55,600 --> 00:26:01,200
estimation it gives you like a density

622
00:25:58,360 --> 00:26:05,520
graph over the x-axis and you can only

623
00:26:01,200 --> 00:26:07,840
do this with 1D graphs but what um this

624
00:26:05,520 --> 00:26:09,960
allows me to see the peak for the train

625
00:26:07,840 --> 00:26:12,159
and the peak for the station and it's

626
00:26:09,960 --> 00:26:14,120
really easy for my computer to figure

627
00:26:12,159 --> 00:26:16,120
out which Peak is which because the

628
00:26:14,120 --> 00:26:20,159
train Peak is really far away from the

629
00:26:16,120 --> 00:26:22,679
origin compared to the station so anyway

630
00:26:20,159 --> 00:26:25,880
I now that this is done and I can find

631
00:26:22,679 --> 00:26:28,320
the slice Woods to use for every frame I

632
00:26:25,880 --> 00:26:31,440
took I went and took out I went and took

633
00:26:28,320 --> 00:26:33,919
some videos of trains

634
00:26:31,440 --> 00:26:35,840
passing so the top right is the one I

635
00:26:33,919 --> 00:26:37,679
took at huntingdale the bottom right is

636
00:26:35,840 --> 00:26:40,039
the one I took at East pacum with my

637
00:26:37,679 --> 00:26:42,279
friend when the station first opened and

638
00:26:40,039 --> 00:26:45,480
the left bottom left is the one that

639
00:26:42,279 --> 00:26:48,480
took at Edis Veil

640
00:26:45,480 --> 00:26:48,480
so

641
00:26:49,880 --> 00:26:59,279
uh yeah so this is this is a picture I

642
00:26:54,000 --> 00:26:59,279
took it's very um yeah

643
00:26:59,720 --> 00:27:03,240
this one of the four pictures I took

644
00:27:01,279 --> 00:27:05,520
it's probably my least favorite but it

645
00:27:03,240 --> 00:27:05,520
still

646
00:27:05,559 --> 00:27:09,760
works if you zoom in a bit there's like

647
00:27:07,919 --> 00:27:12,480
imperfections but you can't zoom in so

648
00:27:09,760 --> 00:27:12,480
you can't see

649
00:27:13,919 --> 00:27:17,919
them this is another one uh this is

650
00:27:16,440 --> 00:27:20,559
probably my second favorite because of

651
00:27:17,919 --> 00:27:24,480
the coloring it has like a really sandy

652
00:27:20,559 --> 00:27:26,360
beach kind of look to it I like the um

653
00:27:24,480 --> 00:27:29,720
the yellow on the blue and it I think it

654
00:27:26,360 --> 00:27:32,960
looks really nice generally um the

655
00:27:29,720 --> 00:27:36,240
second one the next one is funny because

656
00:27:32,960 --> 00:27:38,799
it actually doesn't fit in this slide

657
00:27:36,240 --> 00:27:41,679
um and the cool thing about this

658
00:27:38,799 --> 00:27:44,159
technique is that the slices can like

659
00:27:41,679 --> 00:27:46,480
just tack arbitrarily long so the

660
00:27:44,159 --> 00:27:51,039
pictures can be arbitarily wide and this

661
00:27:46,480 --> 00:27:54,640
picture is actually 156 pixels W

662
00:27:51,039 --> 00:27:57,679
156,000 pixels wide um it actually

663
00:27:54,640 --> 00:28:00,240
lagged out my slideshow uh slideshow

664
00:27:57,679 --> 00:28:03,200
application when I first opened it so I

665
00:28:00,240 --> 00:28:07,840
had to scale it down

666
00:28:03,200 --> 00:28:09,919
um and my last one is my the one I took

667
00:28:07,840 --> 00:28:11,679
at East packin it's probably my favorite

668
00:28:09,919 --> 00:28:13,799
because the shading worked out so well

669
00:28:11,679 --> 00:28:16,760
it was like sort of foggy on the day so

670
00:28:13,799 --> 00:28:18,399
that reflected well as well and if you

671
00:28:16,760 --> 00:28:21,279
want to try this out for yourself it's

672
00:28:18,399 --> 00:28:24,480
all up on GitHub and you can just

673
00:28:21,279 --> 00:28:28,910
download it and try it if if you want um

674
00:28:24,480 --> 00:28:36,720
you might not want to uh and thank you

675
00:28:28,910 --> 00:28:36,720
[Applause]

676
00:28:37,000 --> 00:28:41,720
that was

677
00:28:38,440 --> 00:28:45,799
amazing wow like shiny shiniest of

678
00:28:41,720 --> 00:28:48,799
trains Zach come on up

679
00:28:45,799 --> 00:28:48,799
here

680
00:28:49,080 --> 00:28:56,279
um awesome all right Zach I'm still

681
00:28:53,440 --> 00:28:58,720
thinking about how shiny those trains

682
00:28:56,279 --> 00:29:03,039
were they're so shiny

683
00:28:58,720 --> 00:29:04,679
okay um Zach has been programming for

684
00:29:03,039 --> 00:29:09,360
almost 5

685
00:29:04,679 --> 00:29:11,720
years um a to guess this Emoji I'm going

686
00:29:09,360 --> 00:29:14,720
to

687
00:29:11,720 --> 00:29:14,720
uh

688
00:29:15,679 --> 00:29:24,200
yeah yes thank you we've got an like

689
00:29:19,240 --> 00:29:28,640
Emoji expert in there yeah I know you um

690
00:29:24,200 --> 00:29:31,159
Zach recently found out which might be a

691
00:29:28,640 --> 00:29:33,039
spoil the opposite of a SPO a nice

692
00:29:31,159 --> 00:29:36,760
throwback to something that Renee said

693
00:29:33,039 --> 00:29:39,360
In This Very Room earlier um that llms

694
00:29:36,760 --> 00:29:42,360
might be very smart smart my words not

695
00:29:39,360 --> 00:29:45,320
not his but they cannot count the number

696
00:29:42,360 --> 00:29:49,480
of letters specified in a word for

697
00:29:45,320 --> 00:29:49,480
example how many RS in

698
00:29:49,640 --> 00:29:58,799
Strawberry Hot yeah like show of hands

699
00:29:52,960 --> 00:30:01,840
it's not to not to um

700
00:29:58,799 --> 00:30:01,840
uh here we

701
00:30:01,880 --> 00:30:08,399
go good you might want to make that

702
00:30:04,480 --> 00:30:10,320
window a bit bigger this yeah yep all

703
00:30:08,399 --> 00:30:12,919
right take it away Zach thank you so

704
00:30:10,320 --> 00:30:12,919
much

705
00:30:15,279 --> 00:30:21,240
big uh my name is Zach ett and today

706
00:30:18,760 --> 00:30:24,880
I'll be presenting the pro rocket League

707
00:30:21,240 --> 00:30:26,919
player game Replay YouTube upload a bot

708
00:30:24,880 --> 00:30:28,760
and that's a very long long Twisted name

709
00:30:26,919 --> 00:30:30,679
but I could not think of a better name

710
00:30:28,760 --> 00:30:31,559
so I'll just explain to you quickly what

711
00:30:30,679 --> 00:30:33,880
that

712
00:30:31,559 --> 00:30:35,440
means so first of all I'll just say what

713
00:30:33,880 --> 00:30:37,240
the bot is and also what is rock league

714
00:30:35,440 --> 00:30:38,840
for anyone that does not know uh rocket

715
00:30:37,240 --> 00:30:42,000
league is an online video game that can

716
00:30:38,840 --> 00:30:44,720
be thought of as as car soccer uh it's

717
00:30:42,000 --> 00:30:46,200
normally played in 2v2 or 3v3 formats

718
00:30:44,720 --> 00:30:48,039
and it's very competitive with many

719
00:30:46,200 --> 00:30:50,159
players playing online with the goals of

720
00:30:48,039 --> 00:30:51,600
climbing the rank system uh these

721
00:30:50,159 --> 00:30:53,840
players can watch professional players

722
00:30:51,600 --> 00:30:55,279
play the game online uh and can learn

723
00:30:53,840 --> 00:30:57,320
from them and improve their own game

724
00:30:55,279 --> 00:30:59,200
play so it's quite common to watch

725
00:30:57,320 --> 00:31:02,039
people a lot better than you to try and

726
00:30:59,200 --> 00:31:03,919
improve uh this is where my bot comes in

727
00:31:02,039 --> 00:31:06,519
which hopes to find the replays of

728
00:31:03,919 --> 00:31:09,279
professional rocket league players uh

729
00:31:06,519 --> 00:31:11,720
download those replays and then upload

730
00:31:09,279 --> 00:31:14,200
them onto a YouTube channel for people

731
00:31:11,720 --> 00:31:16,200
to view publicly so they can learn from

732
00:31:14,200 --> 00:31:18,120
them uh and you can see a little

733
00:31:16,200 --> 00:31:20,639
screenshot of like the YouTube channel I

734
00:31:18,120 --> 00:31:24,080
made uh with videos that's got over a th

735
00:31:20,639 --> 00:31:25,440
videos on there um all right so there

736
00:31:24,080 --> 00:31:27,519
are three main parts of the bot first of

737
00:31:25,440 --> 00:31:29,240
all I need to download the replays uh

738
00:31:27,519 --> 00:31:31,399
once the replays are downloaded I need

739
00:31:29,240 --> 00:31:33,960
to physically record them into an MP4

740
00:31:31,399 --> 00:31:36,200
format uh to upload and then the third

741
00:31:33,960 --> 00:31:38,440
part is uploading to

742
00:31:36,200 --> 00:31:40,840
YouTube uh for the first part which is

743
00:31:38,440 --> 00:31:44,159
downloading the replays uh I first of

744
00:31:40,840 --> 00:31:45,880
all had to use an external API or third

745
00:31:44,159 --> 00:31:48,840
party API because rocket League doesn't

746
00:31:45,880 --> 00:31:50,639
have one uh so I use ball chasing. comom

747
00:31:48,840 --> 00:31:53,240
which many people upload their replays

748
00:31:50,639 --> 00:31:56,720
onto automatically using a plugin for

749
00:31:53,240 --> 00:31:58,360
Rocket league so uh there is loads of

750
00:31:56,720 --> 00:32:00,080
different replays on there so it was no

751
00:31:58,360 --> 00:32:01,480
problem downloading replays of lots of

752
00:32:00,080 --> 00:32:06,120
different professional players off of

753
00:32:01,480 --> 00:32:08,440
there um and I had to use that to

754
00:32:06,120 --> 00:32:10,039
download the replays so I first of all

755
00:32:08,440 --> 00:32:12,639
picked a bunch of pro players I wanted

756
00:32:10,039 --> 00:32:14,200
to play and put them into a uh

757
00:32:12,639 --> 00:32:15,559
dictionary so you can see lots of

758
00:32:14,200 --> 00:32:17,519
different Pro players I've selected and

759
00:32:15,559 --> 00:32:22,240
there steam IDs which I need to pass

760
00:32:17,519 --> 00:32:27,639
into the API to download the replays

761
00:32:22,240 --> 00:32:31,000
so oh sorry um once I have the players I

762
00:32:27,639 --> 00:32:33,639
pi a random player using um just picking

763
00:32:31,000 --> 00:32:35,559
a random one and uh put it through the

764
00:32:33,639 --> 00:32:37,600
API which then downloads onto the

765
00:32:35,559 --> 00:32:39,080
computer and I place into a specific

766
00:32:37,600 --> 00:32:41,279
file that rocket League knows to search

767
00:32:39,080 --> 00:32:41,279
for

768
00:32:41,919 --> 00:32:46,200
replays um once the replays downloaded

769
00:32:44,760 --> 00:32:47,960
and is in the folder it's a bit of a

770
00:32:46,200 --> 00:32:49,919
pain to launch because I need to

771
00:32:47,960 --> 00:32:51,399
physically open it in rocket League to

772
00:32:49,919 --> 00:32:54,799
spectate the replay it's not really in

773
00:32:51,399 --> 00:32:57,120
an MP4 format it's more in a special

774
00:32:54,799 --> 00:33:00,399
format that the game just uses to keep

775
00:32:57,120 --> 00:33:03,039
track of input puts and to replay the it

776
00:33:00,399 --> 00:33:04,960
back to you so you need to really watch

777
00:33:03,039 --> 00:33:07,360
the replay in rocket League to then

778
00:33:04,960 --> 00:33:08,840
record it so I need to actually inside

779
00:33:07,360 --> 00:33:11,240
rocket league with rocket League open on

780
00:33:08,840 --> 00:33:13,559
the computer uh it needs to navigate

781
00:33:11,240 --> 00:33:17,440
through the GUI so I use Pi Auto [ __ ]

782
00:33:13,559 --> 00:33:19,200
which is um a emulation sort of uh

783
00:33:17,440 --> 00:33:22,600
Library which lets you simulate key

784
00:33:19,200 --> 00:33:24,559
presses and uh Mouse inputs so you can

785
00:33:22,600 --> 00:33:26,639
see it navigates through the pause menu

786
00:33:24,559 --> 00:33:28,360
which I've got a picture of up on screen

787
00:33:26,639 --> 00:33:29,880
as well as the replay menu and it just

788
00:33:28,360 --> 00:33:33,200
needs to open the replay that it's just

789
00:33:29,880 --> 00:33:35,559
downloaded using the ballchasing

790
00:33:33,200 --> 00:33:37,559
API uh once the replay is open though

791
00:33:35,559 --> 00:33:40,440
there's another big problem it's

792
00:33:37,559 --> 00:33:42,000
spectating a random player so we

793
00:33:40,440 --> 00:33:44,080
obviously want to watch one specific

794
00:33:42,000 --> 00:33:46,240
player that I've picked beforehand so in

795
00:33:44,080 --> 00:33:48,080
this case we're looking at squishy which

796
00:33:46,240 --> 00:33:52,039
is one of the most popular rocket league

797
00:33:48,080 --> 00:33:53,200
players ever uh we need to make sure

798
00:33:52,039 --> 00:33:54,679
we're actually spectating the correct

799
00:33:53,200 --> 00:33:56,480
player though because in this game here

800
00:33:54,679 --> 00:33:58,120
we can see it's a 2v2 game of Rocket

801
00:33:56,480 --> 00:33:59,960
League you can see four players on the

802
00:33:58,120 --> 00:34:01,320
field we don't know necessarily that

803
00:33:59,960 --> 00:34:02,919
we're going to be spectating the correct

804
00:34:01,320 --> 00:34:05,320
player straight away so we need to

805
00:34:02,919 --> 00:34:07,159
somehow figure out how we know what

806
00:34:05,320 --> 00:34:09,280
player we're currently watching and we

807
00:34:07,159 --> 00:34:11,079
need to switch to a different player if

808
00:34:09,280 --> 00:34:14,240
we're not spectating the correct player

809
00:34:11,079 --> 00:34:16,200
so uh you can see currently I've put a

810
00:34:14,240 --> 00:34:18,280
red box around the name of the current

811
00:34:16,200 --> 00:34:20,839
player was spectating so using the py

812
00:34:18,280 --> 00:34:22,679
Tesseract Library you can take a photo

813
00:34:20,839 --> 00:34:25,800
of somewhere on the screen and then pass

814
00:34:22,679 --> 00:34:27,720
it through uh pesera and it basically

815
00:34:25,800 --> 00:34:29,520
extracts text from it into a string form

816
00:34:27,720 --> 00:34:32,000
format which lets you compare the name

817
00:34:29,520 --> 00:34:34,000
with the pro player you want to spectate

818
00:34:32,000 --> 00:34:36,040
so I just cycle through all of the

819
00:34:34,000 --> 00:34:38,440
players

820
00:34:36,040 --> 00:34:40,399
um I don't know if the video is going to

821
00:34:38,440 --> 00:34:42,480
work it never likes working uh when I'm

822
00:34:40,399 --> 00:34:43,879
in the presentat presentator view

823
00:34:42,480 --> 00:34:45,280
there's another later video which we can

824
00:34:43,879 --> 00:34:46,639
watch but it was just going to be a

825
00:34:45,280 --> 00:34:48,159
video that goes through it so we won't

826
00:34:46,639 --> 00:34:49,520
worry about it for the moment uh it

827
00:34:48,159 --> 00:34:50,720
Cycles through all the players finds the

828
00:34:49,520 --> 00:34:52,480
player we want to watch in this case

829
00:34:50,720 --> 00:34:56,119
it's squishy then it's starts the replay

830
00:34:52,480 --> 00:34:56,119
once it's good to go

831
00:34:58,400 --> 00:35:02,200
uh oh y there you go replay working

832
00:35:00,119 --> 00:35:04,440
never mind uh so you can just see it

833
00:35:02,200 --> 00:35:07,320
quickly Cycles through the players and

834
00:35:04,440 --> 00:35:09,280
then picks the player that it thinks

835
00:35:07,320 --> 00:35:10,680
matches closest with the pro player and

836
00:35:09,280 --> 00:35:12,240
then the actual replay begins and you

837
00:35:10,680 --> 00:35:15,440
can see rocket League starting to play

838
00:35:12,240 --> 00:35:15,440
once it's now spectating

839
00:35:16,680 --> 00:35:22,119
squishy uh uh on to the second part

840
00:35:20,119 --> 00:35:23,440
which is recording the replays so as you

841
00:35:22,119 --> 00:35:24,880
can see there we got up to the part

842
00:35:23,440 --> 00:35:26,760
where the actual rocket League game

843
00:35:24,880 --> 00:35:29,760
started so we need to somehow actually

844
00:35:26,760 --> 00:35:33,280
record the replays into an MP4 format uh

845
00:35:29,760 --> 00:35:35,160
to upload onto YouTube so first of all I

846
00:35:33,280 --> 00:35:36,960
I tried out a lot of different routes

847
00:35:35,160 --> 00:35:38,440
for screen recording so I first of all

848
00:35:36,960 --> 00:35:40,560
used open CV because there

849
00:35:38,440 --> 00:35:42,640
screenshotting functionality as well as

850
00:35:40,560 --> 00:35:45,000
an easy ways to pass the screenshots

851
00:35:42,640 --> 00:35:47,040
into video uh but the issue I had was

852
00:35:45,000 --> 00:35:49,640
none of my replays could go above 10 to

853
00:35:47,040 --> 00:35:52,480
20 frames a second and when you upload

854
00:35:49,640 --> 00:35:55,240
videos onto YouTube normally people

855
00:35:52,480 --> 00:35:56,560
upload at 30 to 60 frames a second so no

856
00:35:55,240 --> 00:35:58,960
one would have any interest in really

857
00:35:56,560 --> 00:36:00,880
watching 10 to 20 frame a second videos

858
00:35:58,960 --> 00:36:01,760
otherwise it just looks too choppy uh

859
00:36:00,880 --> 00:36:03,800
especially when you're watching

860
00:36:01,760 --> 00:36:05,800
fast-paced video games so that wasn't

861
00:36:03,800 --> 00:36:07,359
really going to work for me uh I tried

862
00:36:05,800 --> 00:36:10,440
different combinations of libraries I

863
00:36:07,359 --> 00:36:12,480
tried using pi Auto goey again uh to do

864
00:36:10,440 --> 00:36:14,280
screenshots along with open CV a bunch

865
00:36:12,480 --> 00:36:17,040
of different combinations and just

866
00:36:14,280 --> 00:36:19,079
couldn't get it above really 20 FPS

867
00:36:17,040 --> 00:36:20,400
which wasn't good enough uh also a

868
00:36:19,079 --> 00:36:22,800
library that was specifically made for

869
00:36:20,400 --> 00:36:24,599
screen recording called p screen wreck

870
00:36:22,800 --> 00:36:26,359
uh that still was running very slow and

871
00:36:24,599 --> 00:36:27,960
I couldn't get it working so I was a bit

872
00:36:26,359 --> 00:36:30,720
stuck on this and I had idea how to

873
00:36:27,960 --> 00:36:32,119
screen record in Python uh and I just

874
00:36:30,720 --> 00:36:35,200
couldn't really work it out so I had to

875
00:36:32,119 --> 00:36:37,720
try and create something myself um so I

876
00:36:35,200 --> 00:36:39,760
ended up making my own Library uh called

877
00:36:37,720 --> 00:36:42,000
piast screen cap uh which you can

878
00:36:39,760 --> 00:36:43,400
install using pip uh on screen and

879
00:36:42,000 --> 00:36:45,240
there's also all the code available on

880
00:36:43,400 --> 00:36:47,680
GitHub uh it takes advantage of a

881
00:36:45,240 --> 00:36:49,760
certain screenshotting Library called DX

882
00:36:47,680 --> 00:36:52,440
cam which only works on windows so my

883
00:36:49,760 --> 00:36:56,280
library also sadly only works on Windows

884
00:36:52,440 --> 00:36:57,800
um but my library I managed to use and I

885
00:36:56,280 --> 00:37:00,640
managed to get it to record replay is up

886
00:36:57,800 --> 00:37:03,920
to 90 FPS on the computer which is well

887
00:37:00,640 --> 00:37:06,200
above YouTube's maximum of 60 so I I was

888
00:37:03,920 --> 00:37:09,319
able to use that in my application to

889
00:37:06,200 --> 00:37:12,200
then record replays for

890
00:37:09,319 --> 00:37:16,119
me um on to part three uploading the

891
00:37:12,200 --> 00:37:18,960
replays to YouTube so once my uh library

892
00:37:16,119 --> 00:37:21,599
then up records the whole replays I need

893
00:37:18,960 --> 00:37:23,960
to uh physically upload onto re onto

894
00:37:21,599 --> 00:37:27,240
YouTube so it's in the MP4 format ready

895
00:37:23,960 --> 00:37:29,520
to go um so first of all I had a few

896
00:37:27,240 --> 00:37:31,079
options well I thought I had one option

897
00:37:29,520 --> 00:37:33,160
to start with which was using the

898
00:37:31,079 --> 00:37:36,680
YouTube API which I managed to get

899
00:37:33,160 --> 00:37:39,000
working all the way uh to the point that

900
00:37:36,680 --> 00:37:40,599
uploaded the YouTube videos ready to go

901
00:37:39,000 --> 00:37:42,359
uploaded descriptions uploaded

902
00:37:40,599 --> 00:37:45,319
thumbnails everything was working

903
00:37:42,359 --> 00:37:48,480
perfectly until I realized one problem

904
00:37:45,319 --> 00:37:50,960
which was that you cannot upload public

905
00:37:48,480 --> 00:37:52,640
YouTube videos using the API without

906
00:37:50,960 --> 00:37:54,520
getting permission or without filling

907
00:37:52,640 --> 00:37:56,319
out a form for YouTube so that means the

908
00:37:54,520 --> 00:37:59,040
videos although up on the YouTube

909
00:37:56,319 --> 00:38:00,680
platform could only unlisted or private

910
00:37:59,040 --> 00:38:02,200
meaning that no one could publicly see

911
00:38:00,680 --> 00:38:04,920
them and you couldn't even go in and

912
00:38:02,200 --> 00:38:07,319
manually change it uh and I needed to

913
00:38:04,920 --> 00:38:09,839
fill out a form for YouTube to basically

914
00:38:07,319 --> 00:38:11,599
get that changed uh so this meant either

915
00:38:09,839 --> 00:38:16,079
I could spend 30 minutes filling out an

916
00:38:11,599 --> 00:38:18,040
API form um get it approved maybe takes

917
00:38:16,079 --> 00:38:20,599
a couple days but I once again had all

918
00:38:18,040 --> 00:38:21,960
of the code fully working uh wouldn't

919
00:38:20,599 --> 00:38:23,480
have needed to change anything just

920
00:38:21,960 --> 00:38:25,640
needed to fill out the form get it

921
00:38:23,480 --> 00:38:28,079
approved by YouTube or I could spend

922
00:38:25,640 --> 00:38:30,359
many hours uh dele reading all of the

923
00:38:28,079 --> 00:38:31,880
code that I'd just written uh spending

924
00:38:30,359 --> 00:38:33,640
ages thinking of new ideas of how I was

925
00:38:31,880 --> 00:38:36,079
going to do it and eventually coming to

926
00:38:33,640 --> 00:38:39,520
the conclusion I should just automate it

927
00:38:36,079 --> 00:38:42,160
in a similar way to how I automated the

928
00:38:39,520 --> 00:38:44,200
uh rocket League GUI earlier uh by

929
00:38:42,160 --> 00:38:46,319
simply just moving the mouse around the

930
00:38:44,200 --> 00:38:49,240
screen and just emulating keyboard

931
00:38:46,319 --> 00:38:51,920
presses and mouse presses I think you

932
00:38:49,240 --> 00:38:54,800
can figure out which one I went for uh

933
00:38:51,920 --> 00:38:58,160
here's the choice I made so this is just

934
00:38:54,800 --> 00:39:00,040
uh oh yeah here we go

935
00:38:58,160 --> 00:39:02,280
uh we can just watch what the choice I

936
00:39:00,040 --> 00:39:05,880
made

937
00:39:02,280 --> 00:39:08,000
was not working oh there we go uh so

938
00:39:05,880 --> 00:39:10,079
here's the YouTube Studio page uh you

939
00:39:08,000 --> 00:39:12,760
can see the mouse in the top right uh

940
00:39:10,079 --> 00:39:14,960
it's just selecting file open File

941
00:39:12,760 --> 00:39:17,440
Explorer select the

942
00:39:14,960 --> 00:39:19,359
replay going through waits for the video

943
00:39:17,440 --> 00:39:21,400
to pause this section could have taken a

944
00:39:19,359 --> 00:39:23,079
while so it pauses here for a second so

945
00:39:21,400 --> 00:39:25,520
just give it a couple seconds then it

946
00:39:23,079 --> 00:39:27,800
starts pasting in the title doing all

947
00:39:25,520 --> 00:39:29,800
keyboard presses pasting the information

948
00:39:27,800 --> 00:39:30,680
based on the replay does the same for

949
00:39:29,800 --> 00:39:34,280
the

950
00:39:30,680 --> 00:39:36,119
description uh then Scrolls down need to

951
00:39:34,280 --> 00:39:37,920
say the video is not made for kids so

952
00:39:36,119 --> 00:39:40,960
just check that box really quickly and

953
00:39:37,920 --> 00:39:43,440
show more scroll

954
00:39:40,960 --> 00:39:45,200
down adds in some tags which just helps

955
00:39:43,440 --> 00:39:47,160
with the YouTube

956
00:39:45,200 --> 00:39:49,200
discoverability uh p in a bunch of

957
00:39:47,160 --> 00:39:51,680
different things

958
00:39:49,200 --> 00:39:53,359
there uh click enter a few times and

959
00:39:51,680 --> 00:39:55,640
here we go public video which I couldn't

960
00:39:53,359 --> 00:39:57,400
get before with the API also copies the

961
00:39:55,640 --> 00:39:59,200
video link Just for future thumbnail

962
00:39:57,400 --> 00:40:01,800
uploading which I still actually did use

963
00:39:59,200 --> 00:40:04,190
the YouTube API for uh that's how it

964
00:40:01,800 --> 00:40:09,520
uploads the videos

965
00:40:04,190 --> 00:40:13,440
[Applause]

966
00:40:09,520 --> 00:40:16,960
um uh Frozen that's not

967
00:40:13,440 --> 00:40:20,119
good oh

968
00:40:16,960 --> 00:40:21,680
oh oh there we go um all right uh

969
00:40:20,119 --> 00:40:22,680
there's an other mention though sort of

970
00:40:21,680 --> 00:40:23,920
wanted to make I don't think there's

971
00:40:22,680 --> 00:40:25,800
enough time to talk about it in this

972
00:40:23,920 --> 00:40:28,000
slide but I also need to generate the

973
00:40:25,800 --> 00:40:29,960
thumbnail so he can see an example

974
00:40:28,000 --> 00:40:32,119
thumbnail of one of the YouTube videos

975
00:40:29,960 --> 00:40:35,000
where it uh picks a random car and it

976
00:40:32,119 --> 00:40:36,720
picks a random field and enters in some

977
00:40:35,000 --> 00:40:39,480
of the information based on the replay

978
00:40:36,720 --> 00:40:42,800
so that's an example replay of the pro

979
00:40:39,480 --> 00:40:44,119
player Zen um that's also done in python

980
00:40:42,800 --> 00:40:46,880
as well randomly generating the

981
00:40:44,119 --> 00:40:49,520
thumbnails but uh I think that's a bit

982
00:40:46,880 --> 00:40:52,280
too much to talk about today um lastly

983
00:40:49,520 --> 00:40:56,200
was the project a success

984
00:40:52,280 --> 00:40:57,960
um I left the bot running 24/7 between

985
00:40:56,200 --> 00:41:00,119
mid August to Mid setember last year so

986
00:40:57,960 --> 00:41:02,440
it was running for about 4 weeks uh

987
00:41:00,119 --> 00:41:04,000
straight in towards the end of last year

988
00:41:02,440 --> 00:41:06,000
and it managed to upload over a th000

989
00:41:04,000 --> 00:41:08,720
videos onto the YouTube channel uh

990
00:41:06,000 --> 00:41:10,200
without me interrupting it uh so it was

991
00:41:08,720 --> 00:41:12,640
successful in the fact that it did its

992
00:41:10,200 --> 00:41:14,960
job although it wasn't necessarily the

993
00:41:12,640 --> 00:41:18,800
most popular uh across that month it

994
00:41:14,960 --> 00:41:20,240
gained about 814 views total uh which

995
00:41:18,800 --> 00:41:21,599
like still quite a lot for me doing

996
00:41:20,240 --> 00:41:24,400
absolutely nothing and just leaving it

997
00:41:21,599 --> 00:41:26,680
running but uh that meant like it was

998
00:41:24,400 --> 00:41:28,040
averaging less than a viewer videos so

999
00:41:26,680 --> 00:41:30,400
there were quite a few videos that

1000
00:41:28,040 --> 00:41:33,240
reached probably in the T like probably

1001
00:41:30,400 --> 00:41:35,119
30 40 views but probably 80% of those

1002
00:41:33,240 --> 00:41:37,960
videos didn't even get a single view so

1003
00:41:35,119 --> 00:41:39,760
it was quite successful uh in doing its

1004
00:41:37,960 --> 00:41:42,480
job but not quite successful in many

1005
00:41:39,760 --> 00:41:44,880
people weren't really watching it um

1006
00:41:42,480 --> 00:41:44,880
thank you for

1007
00:41:50,079 --> 00:41:54,640
listening amazing four subscribers

1008
00:41:53,000 --> 00:41:56,040
though four subscribers I think you

1009
00:41:54,640 --> 00:41:58,720
reached five at one point until

1010
00:41:56,040 --> 00:42:00,280
subscribed

1011
00:41:58,720 --> 00:42:01,839
you got to wonder what's happening in

1012
00:42:00,280 --> 00:42:06,000
that that video to make someone

1013
00:42:01,839 --> 00:42:10,319
unsubscribe yeah I thank you so much

1014
00:42:06,000 --> 00:42:13,400
next up we have Annabelle Annabelle is

1015
00:42:10,319 --> 00:42:14,880
has been programming for three years oh

1016
00:42:13,400 --> 00:42:17,839
gosh hang

1017
00:42:14,880 --> 00:42:22,680
on uh this is going to be tricky all

1018
00:42:17,839 --> 00:42:22,680
right this Emoji depending yeah

1019
00:42:25,319 --> 00:42:30,160
it yeah I if if I were green and a lot

1020
00:42:28,480 --> 00:42:32,280
more

1021
00:42:30,160 --> 00:42:36,119
flexible

1022
00:42:32,280 --> 00:42:36,119
um yeah

1023
00:42:36,240 --> 00:42:41,079
Dragon I was really worried about that

1024
00:42:38,400 --> 00:42:42,319
one okay uh Annabelle's favorite emoji

1025
00:42:41,079 --> 00:42:46,240
is the dragon

1026
00:42:42,319 --> 00:42:48,040
Emoji um and the thing that she recently

1027
00:42:46,240 --> 00:42:51,359
learned and wanted to share with us

1028
00:42:48,040 --> 00:42:54,960
about computers which I tested out today

1029
00:42:51,359 --> 00:42:58,640
and can confirm works is that if you

1030
00:42:54,960 --> 00:43:01,040
Google best font ever

1031
00:42:58,640 --> 00:43:04,559
the font of Google changes to does

1032
00:43:01,040 --> 00:43:08,119
anyone want to guess which font comic

1033
00:43:04,559 --> 00:43:10,920
Sands beautiful uh I happen to test this

1034
00:43:08,119 --> 00:43:17,280
while also typoing and it still

1035
00:43:10,920 --> 00:43:19,440
worked so yes good uh good work there um

1036
00:43:17,280 --> 00:43:23,359
are we good

1037
00:43:19,440 --> 00:43:28,880
fantastic um okay can we give Annabelle

1038
00:43:23,359 --> 00:43:28,880
a big round of applause take it away

1039
00:43:30,119 --> 00:43:33,599
hi I'm Annabelle I'm a 16 years old and

1040
00:43:32,280 --> 00:43:35,920
I'm a recent graduate of glenunga

1041
00:43:33,599 --> 00:43:37,920
International High School from adalade

1042
00:43:35,920 --> 00:43:39,359
um I've been programming for three years

1043
00:43:37,920 --> 00:43:40,599
and today I'll be talking about my

1044
00:43:39,359 --> 00:43:43,440
project

1045
00:43:40,599 --> 00:43:45,000
Pythagoras so what is Pythagoras it's

1046
00:43:43,440 --> 00:43:46,760
kind of hard to explain so I'll let this

1047
00:43:45,000 --> 00:43:49,160
demonstrate here's a screenshot of a

1048
00:43:46,760 --> 00:43:50,760
video game Celeste the program analyzes

1049
00:43:49,160 --> 00:43:53,200
the image to determine which colors are

1050
00:43:50,760 --> 00:43:55,440
most similar to the provided SVG that

1051
00:43:53,200 --> 00:43:56,720
pride flag then replaces the colors

1052
00:43:55,440 --> 00:43:58,720
producing the pride flag you see on the

1053
00:43:56,720 --> 00:44:00,720
right hand inside essentially you get a

1054
00:43:58,720 --> 00:44:02,800
nice SVG using the same plette of your

1055
00:44:00,720 --> 00:44:04,640
image more technically it is a

1056
00:44:02,800 --> 00:44:06,680
multiplatform command line interface

1057
00:44:04,640 --> 00:44:08,520
program as well as an installable python

1058
00:44:06,680 --> 00:44:10,839
package all of which can be run entirely

1059
00:44:08,520 --> 00:44:12,599
in your terminal the original reason I

1060
00:44:10,839 --> 00:44:14,640
created it as seen here was to generate

1061
00:44:12,599 --> 00:44:16,440
Pride flags from the predominant colors

1062
00:44:14,640 --> 00:44:18,760
in a photo which leads me to my next

1063
00:44:16,440 --> 00:44:20,559
Point um you may now be thinking what

1064
00:44:18,760 --> 00:44:24,440
the hell am I going to use this for and

1065
00:44:20,559 --> 00:44:26,640
that's totally fair I don't know

1066
00:44:24,440 --> 00:44:28,800
anyways here's a short overview of the

1067
00:44:26,640 --> 00:44:30,880
process used to develop Pythagoras I'll

1068
00:44:28,800 --> 00:44:33,200
talk about a few of them here today um

1069
00:44:30,880 --> 00:44:35,480
using regular expressions numpy and

1070
00:44:33,200 --> 00:44:38,599
pillow to process images and extract

1071
00:44:35,480 --> 00:44:39,960
colors um determining similar colors

1072
00:44:38,599 --> 00:44:41,920
I'll talk about that a bit later

1073
00:44:39,960 --> 00:44:44,559
substituting colors into the final SVG

1074
00:44:41,920 --> 00:44:47,839
and then generating it and producing it

1075
00:44:44,559 --> 00:44:49,640
for the user to use so the first step is

1076
00:44:47,839 --> 00:44:51,920
to obtain a list of the colors from The

1077
00:44:49,640 --> 00:44:54,040
Source image that we can work with as a

1078
00:44:51,920 --> 00:44:56,040
command line program I used ARG paaths

1079
00:44:54,040 --> 00:44:58,240
to interpret the user's input to get the

1080
00:44:56,040 --> 00:45:00,240
path to the image then pillows very

1081
00:44:58,240 --> 00:45:01,440
convenient get colors function to obtain

1082
00:45:00,240 --> 00:45:03,880
a list of

1083
00:45:01,440 --> 00:45:06,559
colors next we do the same thing but

1084
00:45:03,880 --> 00:45:08,480
with SVG svgs are a type of vector image

1085
00:45:06,559 --> 00:45:11,040
format that can be scaled infinitely

1086
00:45:08,480 --> 00:45:12,720
while still um preserving quality so

1087
00:45:11,040 --> 00:45:15,760
they're fairly popular for representing

1088
00:45:12,720 --> 00:45:17,760
icons and in my case flags colors are

1089
00:45:15,760 --> 00:45:19,200
present in svgs as hex codes so the

1090
00:45:17,760 --> 00:45:21,960
question is how do we find all those hex

1091
00:45:19,200 --> 00:45:25,520
codes the answer is unfortunately

1092
00:45:21,960 --> 00:45:27,440
Rex so hex hex codes as we usually know

1093
00:45:25,520 --> 00:45:28,880
them generally take the same form they

1094
00:45:27,440 --> 00:45:30,599
start with a hashtag and they have six

1095
00:45:28,880 --> 00:45:32,400
characters after them so it should be

1096
00:45:30,599 --> 00:45:36,839
pretty easy to find and extract a hex

1097
00:45:32,400 --> 00:45:36,839
code using a regular expression such as

1098
00:45:37,040 --> 00:45:41,359
this so here's a reminder of the Rex I

1099
00:45:39,960 --> 00:45:44,400
developed from the previous slide it

1100
00:45:41,359 --> 00:45:47,000
looks great so let's test this out so

1101
00:45:44,400 --> 00:45:49,359
here we have a have some code for an SVG

1102
00:45:47,000 --> 00:45:52,400
that's filled with a a shade of magenta

1103
00:45:49,359 --> 00:45:54,119
so that one matches great okay so it

1104
00:45:52,400 --> 00:45:55,839
doesn't seem to work with three digit

1105
00:45:54,119 --> 00:45:58,040
hex codes that's because we only check

1106
00:45:55,839 --> 00:46:00,520
for six and

1107
00:45:58,040 --> 00:46:02,720
finally I failed to account for the fact

1108
00:46:00,520 --> 00:46:06,119
that SVG support arbitrary color names

1109
00:46:02,720 --> 00:46:10,559
as fill colors including Papa whip which

1110
00:46:06,119 --> 00:46:10,559
is quite frankly a nonsense name for a

1111
00:46:10,720 --> 00:46:15,319
color so now you need to remember that

1112
00:46:13,520 --> 00:46:16,880
at the time I was making this it was

1113
00:46:15,319 --> 00:46:19,200
originally intended to only be used for

1114
00:46:16,880 --> 00:46:22,040
Pride Flags so I seeded the program with

1115
00:46:19,200 --> 00:46:24,520
an assortment of pride flag SVG files

1116
00:46:22,040 --> 00:46:26,200
that users could use by default this of

1117
00:46:24,520 --> 00:46:28,720
course meant that I had to check if all

1118
00:46:26,200 --> 00:46:30,920
the colors in in my template svgs would

1119
00:46:28,720 --> 00:46:32,440
actually work with my program and they

1120
00:46:30,920 --> 00:46:34,119
weren't affected by any of the three

1121
00:46:32,440 --> 00:46:36,200
issues the issues I saw on the previous

1122
00:46:34,119 --> 00:46:38,160
slide so I thought to myself I'm a

1123
00:46:36,200 --> 00:46:40,920
programmer I can make a quick program to

1124
00:46:38,160 --> 00:46:42,960
like sort that all out instead of being

1125
00:46:40,920 --> 00:46:46,280
boring and reading through all my SVG

1126
00:46:42,960 --> 00:46:47,839
files to exchange them manually so I

1127
00:46:46,280 --> 00:46:49,800
went through each file and exchange it

1128
00:46:47,839 --> 00:46:52,520
man exchange it man manually and you can

1129
00:46:49,800 --> 00:46:55,960
see there bunch of files I changed um

1130
00:46:52,520 --> 00:46:57,640
took me over an hour don't do it so

1131
00:46:55,960 --> 00:46:59,680
later on I went ahead and wrote a

1132
00:46:57,640 --> 00:47:02,359
function this converts all three-digit

1133
00:46:59,680 --> 00:47:05,079
hex codes into six-digit hex codes

1134
00:47:02,359 --> 00:47:07,440
converts um nonsense color names into

1135
00:47:05,079 --> 00:47:09,880
actual hex codes by reading from a Json

1136
00:47:07,440 --> 00:47:12,480
file so it standard basically

1137
00:47:09,880 --> 00:47:15,440
standardizes the svgs to be used

1138
00:47:12,480 --> 00:47:17,280
normally okay next the reason why I

1139
00:47:15,440 --> 00:47:19,839
wanted to extract the hex codes of each

1140
00:47:17,280 --> 00:47:22,200
color is because it's easier to find

1141
00:47:19,839 --> 00:47:24,599
similar colors this way colors can also

1142
00:47:22,200 --> 00:47:26,160
be expressed in an RGB format we can

1143
00:47:24,599 --> 00:47:27,920
show this as a topple consisting of

1144
00:47:26,160 --> 00:47:29,760
three elements the red component the

1145
00:47:27,920 --> 00:47:31,640
green component and the blue component

1146
00:47:29,760 --> 00:47:33,839
if we think of these colors as points in

1147
00:47:31,640 --> 00:47:35,440
3D Vector space then the similarity of

1148
00:47:33,839 --> 00:47:37,520
two colors can be found by determining

1149
00:47:35,440 --> 00:47:40,119
the shortest distance between them or

1150
00:47:37,520 --> 00:47:42,760
like the modulus of the vector between

1151
00:47:40,119 --> 00:47:44,960
them here I use Python's math module to

1152
00:47:42,760 --> 00:47:47,559
find the linear distance between any two

1153
00:47:44,960 --> 00:47:50,079
colors then ultimately identify which

1154
00:47:47,559 --> 00:47:52,680
pairing produced the minimum

1155
00:47:50,079 --> 00:47:54,680
difference theoretically this works the

1156
00:47:52,680 --> 00:47:56,520
problem is that colors that the program

1157
00:47:54,680 --> 00:47:58,440
deems are similar aren't really what we

1158
00:47:56,520 --> 00:48:00,240
would consider the same for instance

1159
00:47:58,440 --> 00:48:01,920
it's generally agreed upon that the

1160
00:48:00,240 --> 00:48:04,040
human eye can distinguish more shades of

1161
00:48:01,920 --> 00:48:06,200
green than anything else and the way

1162
00:48:04,040 --> 00:48:07,800
that each individual person um sees

1163
00:48:06,200 --> 00:48:09,680
color is always going to be different

1164
00:48:07,800 --> 00:48:12,130
even just by a small amount here's an

1165
00:48:09,680 --> 00:48:14,119
example I'm sure many of you

1166
00:48:12,130 --> 00:48:16,200
[Music]

1167
00:48:14,119 --> 00:48:18,079
recognize so one night when searching

1168
00:48:16,200 --> 00:48:19,839
for a solution to this issue I stumbled

1169
00:48:18,079 --> 00:48:21,200
upon this website which provided handy

1170
00:48:19,839 --> 00:48:23,240
conversion for colors into different

1171
00:48:21,200 --> 00:48:25,119
color spaces in particular there's

1172
00:48:23,240 --> 00:48:27,040
something called the cab color space

1173
00:48:25,119 --> 00:48:29,319
which aims to accurately replicate how

1174
00:48:27,040 --> 00:48:31,599
humans observe color by recognizing that

1175
00:48:29,319 --> 00:48:33,440
it's really actually quite nonlinear I

1176
00:48:31,599 --> 00:48:36,240
added a python implementation to convert

1177
00:48:33,440 --> 00:48:38,400
an RGB tle into a cab topple as an

1178
00:48:36,240 --> 00:48:40,559
alternative algorithm to just straight

1179
00:48:38,400 --> 00:48:43,079
Pythagoras's Theorem this generally

1180
00:48:40,559 --> 00:48:44,839
produces more accurate um results in

1181
00:48:43,079 --> 00:48:46,920
terms of color perception but it isn't

1182
00:48:44,839 --> 00:48:49,280
always the

1183
00:48:46,920 --> 00:48:51,160
case Okay a lot of things about this

1184
00:48:49,280 --> 00:48:52,599
project are subjective for instance how

1185
00:48:51,160 --> 00:48:55,160
do you know that the colors the program

1186
00:48:52,599 --> 00:48:58,240
spits out are actually like from my

1187
00:48:55,160 --> 00:48:59,920
image there are two parts of this first

1188
00:48:58,240 --> 00:49:02,359
you log the coordinates of where the

1189
00:48:59,920 --> 00:49:04,599
colors are found numpy lets you do this

1190
00:49:02,359 --> 00:49:06,200
provided you pass it an image you are

1191
00:49:04,599 --> 00:49:08,720
then able to retrieve a two-dimensional

1192
00:49:06,200 --> 00:49:10,839
array where each cell is equivalent to a

1193
00:49:08,720 --> 00:49:12,400
pixel correspondingly each cell

1194
00:49:10,839 --> 00:49:14,480
represents a color from the original

1195
00:49:12,400 --> 00:49:16,440
image all that's left to do is filter

1196
00:49:14,480 --> 00:49:18,520
for the colors that we want then get the

1197
00:49:16,440 --> 00:49:20,079
coordinates from there secondly you have

1198
00:49:18,520 --> 00:49:21,760
to actually actually find a way to

1199
00:49:20,079 --> 00:49:22,839
represent the coordinates on the image

1200
00:49:21,760 --> 00:49:25,280
which is something that we can do with

1201
00:49:22,839 --> 00:49:27,079
pillow I wrote a function to dynamically

1202
00:49:25,280 --> 00:49:28,960
size and draw a surrounding Circle based

1203
00:49:27,079 --> 00:49:30,880
on the size of the original image so if

1204
00:49:28,960 --> 00:49:33,359
you enable the Highlight flag a new

1205
00:49:30,880 --> 00:49:35,079
image will be produced with red circles

1206
00:49:33,359 --> 00:49:37,359
uh you can probably see a couple over

1207
00:49:35,079 --> 00:49:40,160
there um showing where each of the

1208
00:49:37,359 --> 00:49:42,079
similar colors can be found so yeah just

1209
00:49:40,160 --> 00:49:45,240
check over there and each of those red

1210
00:49:42,079 --> 00:49:47,400
circles corresponds to one color on the

1211
00:49:45,240 --> 00:49:50,000
final

1212
00:49:47,400 --> 00:49:51,640
SVG okay I'll be showing a few other

1213
00:49:50,000 --> 00:49:54,920
examples here with brand logos that

1214
00:49:51,640 --> 00:49:56,880
you're probably familiar with um here's

1215
00:49:54,920 --> 00:49:59,119
a quick example using a screenshot my

1216
00:49:56,880 --> 00:50:01,359
terminal and the slack logo using the

1217
00:49:59,119 --> 00:50:04,400
cab algorithm so using it you can see it

1218
00:50:01,359 --> 00:50:06,680
produces a more muted

1219
00:50:04,400 --> 00:50:08,760
color and here's another one that uses a

1220
00:50:06,680 --> 00:50:11,200
picture of alen snakes live bag and the

1221
00:50:08,760 --> 00:50:12,599
Google logo with a low cost approxim

1222
00:50:11,200 --> 00:50:15,599
approximation this is a different

1223
00:50:12,599 --> 00:50:16,920
algorithm um the not noticeable change

1224
00:50:15,599 --> 00:50:18,880
here is that the program has found a

1225
00:50:16,920 --> 00:50:21,280
lighter blue on the source image to be

1226
00:50:18,880 --> 00:50:24,720
most similar to the original

1227
00:50:21,280 --> 00:50:28,040
color um I've got a video demo here but

1228
00:50:24,720 --> 00:50:31,400
I can do a demo uh like a actual live

1229
00:50:28,040 --> 00:50:31,400
demo if that

1230
00:50:33,000 --> 00:50:36,000
works

1231
00:50:36,240 --> 00:50:42,920
okay open my terminal over here feel you

1232
00:50:39,480 --> 00:50:42,920
can see that

1233
00:51:03,119 --> 00:51:08,640
okay that what I feel so um you can see

1234
00:51:06,480 --> 00:51:11,640
there are a couple of things in that um

1235
00:51:08,640 --> 00:51:16,920
pan line aduino PNG represents my source

1236
00:51:11,640 --> 00:51:19,680
image um- F Pyon SVG so I've decided to

1237
00:51:16,920 --> 00:51:21,960
uh see if I can rep replicate the Pyon

1238
00:51:19,680 --> 00:51:24,559
um logo using my image source image

1239
00:51:21,960 --> 00:51:26,680
Arduino PNG and there are two flags

1240
00:51:24,559 --> 00:51:29,160
verose and highlight highlight will

1241
00:51:26,680 --> 00:51:32,119
produce the red circles and verose will

1242
00:51:29,160 --> 00:51:32,119
just log a bunch of

1243
00:51:33,440 --> 00:51:40,200
stuff okay hopefully that

1244
00:51:36,400 --> 00:51:40,200
works yeah so there we

1245
00:51:40,839 --> 00:51:44,119
go and

1246
00:51:44,559 --> 00:51:50,000
now I will open the actual image

1247
00:51:57,119 --> 00:51:59,440
okay

1248
00:52:01,400 --> 00:52:09,520
so don't know if that's opened open

1249
00:52:06,559 --> 00:52:13,079
yourt oh it's opened here

1250
00:52:09,520 --> 00:52:16,000
yeah that is the pon logo that it's

1251
00:52:13,079 --> 00:52:17,480
disappeared again that's the pon logo um

1252
00:52:16,000 --> 00:52:20,680
generated

1253
00:52:17,480 --> 00:52:23,079
from this random picture of an Arduino I

1254
00:52:20,680 --> 00:52:23,079
had lying

1255
00:52:23,160 --> 00:52:29,240
around and if you look at the logo logo

1256
00:52:27,160 --> 00:52:31,240
up here that's the original one and this

1257
00:52:29,240 --> 00:52:33,640
is my new one and this one you can Al

1258
00:52:31,240 --> 00:52:34,440
also see is kind of less saturated than

1259
00:52:33,640 --> 00:52:41,280
the

1260
00:52:34,440 --> 00:52:41,280
original okay um now go back to our

1261
00:52:42,720 --> 00:52:45,720
presentation

1262
00:52:46,160 --> 00:52:50,480
yeah okay so a couple other things about

1263
00:52:48,680 --> 00:52:52,200
this program um I've documented a lot of

1264
00:52:50,480 --> 00:52:54,079
it online and also provided tools for

1265
00:52:52,200 --> 00:52:56,079
debugging and development in an open

1266
00:52:54,079 --> 00:52:58,720
source repository um there are a few

1267
00:52:56,079 --> 00:53:01,160
floors however and that's uh it's quite

1268
00:52:58,720 --> 00:53:03,079
slow when you run it with large um large

1269
00:53:01,160 --> 00:53:05,040
images high resolution images and

1270
00:53:03,079 --> 00:53:07,440
secondly it favors exact color matches

1271
00:53:05,040 --> 00:53:09,160
over the dominant pette of the image so

1272
00:53:07,440 --> 00:53:11,280
it's easier for the program to find

1273
00:53:09,160 --> 00:53:13,520
exact color matches but they don't fully

1274
00:53:11,280 --> 00:53:16,359
like represent the image um lastly it

1275
00:53:13,520 --> 00:53:17,920
doesn't support some more complex svgs I

1276
00:53:16,359 --> 00:53:21,440
haven't looked a bit too much into that

1277
00:53:17,920 --> 00:53:25,559
because I only found about it last night

1278
00:53:21,440 --> 00:53:27,920
and so yeah does it work technically it

1279
00:53:25,559 --> 00:53:30,319
does and you can try it out by typing uh

1280
00:53:27,920 --> 00:53:32,640
pip install Pythagoras wherever you want

1281
00:53:30,319 --> 00:53:34,760
um haven't been haven't tested it on Mac

1282
00:53:32,640 --> 00:53:36,160
though uh so over the course of this

1283
00:53:34,760 --> 00:53:40,079
presentation did I think of any more

1284
00:53:36,160 --> 00:53:42,960
uses for it uh no uh it's essentially a

1285
00:53:40,079 --> 00:53:45,599
novelty tool but did I have fun making

1286
00:53:42,960 --> 00:53:45,599
it

1287
00:53:45,640 --> 00:53:50,240
absolutely

1288
00:53:48,319 --> 00:53:54,640
yeah thank you for

1289
00:53:50,240 --> 00:53:59,000
[Applause]

1290
00:53:54,640 --> 00:54:02,319
listening thank you so much I I um I

1291
00:53:59,000 --> 00:54:03,720
needed to pause and take a photo of that

1292
00:54:02,319 --> 00:54:06,119
uh yeah come on come on Deck come on

1293
00:54:03,720 --> 00:54:07,920
deck of your last de because I'm going

1294
00:54:06,119 --> 00:54:11,280
to be making sure that those links are

1295
00:54:07,920 --> 00:54:12,880
posted in the education track on Discord

1296
00:54:11,280 --> 00:54:16,440
uh because you may not have thought of

1297
00:54:12,880 --> 00:54:19,640
use for this but I guarantee you that in

1298
00:54:16,440 --> 00:54:22,640
the audience you have a lot of people uh

1299
00:54:19,640 --> 00:54:26,079
trying to find and finding good uses for

1300
00:54:22,640 --> 00:54:30,200
this thank you so much next up the

1301
00:54:26,079 --> 00:54:31,720
second last talk is Toby Toby has been

1302
00:54:30,200 --> 00:54:37,920
programming

1303
00:54:31,720 --> 00:54:40,280
for 5 to six years um oh gosh okay his

1304
00:54:37,920 --> 00:54:40,280
favorite

1305
00:54:42,200 --> 00:54:50,200
emoji eyes eyes yes okay

1306
00:54:46,440 --> 00:54:51,920
yeah um and recently he made the mistake

1307
00:54:50,200 --> 00:54:54,559
of learning about pointers and now they

1308
00:54:51,920 --> 00:54:56,680
haunt your

1309
00:54:54,559 --> 00:54:59,920
nightmares we've all been there we've

1310
00:54:56,680 --> 00:55:02,359
all been there Tak it away Toby hello um

1311
00:54:59,920 --> 00:55:04,760
I am Toby and I really like noise

1312
00:55:02,359 --> 00:55:06,680
algorithms I find them really cool the

1313
00:55:04,760 --> 00:55:09,280
fact that a computer can take in some

1314
00:55:06,680 --> 00:55:11,079
random inputs in a vector and return

1315
00:55:09,280 --> 00:55:12,839
value that's like kind of like its

1316
00:55:11,079 --> 00:55:13,799
neighbors but kind of different is

1317
00:55:12,839 --> 00:55:16,319
really

1318
00:55:13,799 --> 00:55:18,440
fascinating um but what is a noise

1319
00:55:16,319 --> 00:55:20,559
function a noise is the name for the

1320
00:55:18,440 --> 00:55:22,559
output from a noise function and will

1321
00:55:20,559 --> 00:55:24,599
typically be a value between 0 and 1 or

1322
00:55:22,559 --> 00:55:27,079
0 and

1323
00:55:24,599 --> 00:55:30,480
255 it can take several forms it can be

1324
00:55:27,079 --> 00:55:32,400
2D 3D 1D or even 4D depending on the

1325
00:55:30,480 --> 00:55:34,520
function many noise functions can

1326
00:55:32,400 --> 00:55:37,440
actually be fairly easily adapted to

1327
00:55:34,520 --> 00:55:39,520
work in any Dimension n uh but yeah

1328
00:55:37,440 --> 00:55:41,960
that's the main three that noise

1329
00:55:39,520 --> 00:55:44,119
functions normally work in uh here's

1330
00:55:41,960 --> 00:55:45,839
some examples of noise functions so a

1331
00:55:44,119 --> 00:55:47,520
lot of noise functions will output

1332
00:55:45,839 --> 00:55:49,839
complex gradients like the ones I'm

1333
00:55:47,520 --> 00:55:51,680
showing now they've got a really large

1334
00:55:49,839 --> 00:55:53,720
amount of use cases I've personally seen

1335
00:55:51,680 --> 00:55:55,280
them be used for train generation where

1336
00:55:53,720 --> 00:55:57,680
the values returned will correspond to

1337
00:55:55,280 --> 00:55:59,520
Heights so we've got pein noise which is

1338
00:55:57,680 --> 00:56:01,720
some fancy product of vector dot

1339
00:55:59,520 --> 00:56:03,920
products uh then fractal pear noise

1340
00:56:01,720 --> 00:56:06,160
where you layer multiple ones Veron

1341
00:56:03,920 --> 00:56:08,559
noise where you get uh random points and

1342
00:56:06,160 --> 00:56:10,839
find the distance to them Simplex noise

1343
00:56:08,559 --> 00:56:13,400
which I have no idea and my one which

1344
00:56:10,839 --> 00:56:15,799
has been colored fractal iterative Point

1345
00:56:13,400 --> 00:56:18,760
averaging um so this sort of thing can

1346
00:56:15,799 --> 00:56:22,280
be seen in games like Minecraft which is

1347
00:56:18,760 --> 00:56:24,319
a 3D sandbox and Terraria a 2d sandbox

1348
00:56:22,280 --> 00:56:26,480
both of these games rely really heavily

1349
00:56:24,319 --> 00:56:29,039
on noise functions to help create the

1350
00:56:26,480 --> 00:56:31,880
random and in Minecraft's case infinite

1351
00:56:29,039 --> 00:56:34,119
or near infinite uh terrain

1352
00:56:31,880 --> 00:56:37,000
generation noise can also be used for

1353
00:56:34,119 --> 00:56:38,880
animated textures and CGI in fact one of

1354
00:56:37,000 --> 00:56:40,920
the first noise functions invented was

1355
00:56:38,880 --> 00:56:42,760
pein noise and it was invented for the

1356
00:56:40,920 --> 00:56:44,680
movie Tron where it was used to create

1357
00:56:42,760 --> 00:56:47,240
land masses cycling textures and

1358
00:56:44,680 --> 00:56:49,680
probably way more but uh it's hard to

1359
00:56:47,240 --> 00:56:52,920
find specific use cases for it in

1360
00:56:49,680 --> 00:56:54,400
it uh I came up with a basic idea behind

1361
00:56:52,920 --> 00:56:57,480
the noise algorithm that I'll be talking

1362
00:56:54,400 --> 00:57:00,319
about back in like April of

1363
00:56:57,480 --> 00:57:02,400
2023 I think and I was still just

1364
00:57:00,319 --> 00:57:03,799
getting the hang of python back then so

1365
00:57:02,400 --> 00:57:06,079
I made the first version of the

1366
00:57:03,799 --> 00:57:07,640
algorithm in scratch and I was going to

1367
00:57:06,079 --> 00:57:10,640
talk about that but it was like 12

1368
00:57:07,640 --> 00:57:12,039
minutes so I cut that bit out um it's

1369
00:57:10,640 --> 00:57:13,799
actually a really simple algorithm

1370
00:57:12,039 --> 00:57:16,559
though uh you start by placing any

1371
00:57:13,799 --> 00:57:18,079
amount of points on like a pixel canvas

1372
00:57:16,559 --> 00:57:20,960
each with a completely random value

1373
00:57:18,079 --> 00:57:22,760
assigned between 0 and 255 next you get

1374
00:57:20,960 --> 00:57:25,640
a list of all the possible coordinates

1375
00:57:22,760 --> 00:57:26,799
shuffle them and then go through uh it

1376
00:57:25,640 --> 00:57:28,359
one by one

1377
00:57:26,799 --> 00:57:30,160
for each element of that list you place

1378
00:57:28,359 --> 00:57:31,680
a point at the specified coordinates

1379
00:57:30,160 --> 00:57:33,839
then Loop over all the points to find

1380
00:57:31,680 --> 00:57:35,400
the two closest points to it the new

1381
00:57:33,839 --> 00:57:37,640
points value is set to the average of

1382
00:57:35,400 --> 00:57:39,720
the two closest points and once that's

1383
00:57:37,640 --> 00:57:41,240
done you move on to the next once you've

1384
00:57:39,720 --> 00:57:42,640
done this for all the points in the list

1385
00:57:41,240 --> 00:57:44,079
you're able to step back and view the

1386
00:57:42,640 --> 00:57:46,920
finished

1387
00:57:44,079 --> 00:57:49,200
texture now the only problem with this

1388
00:57:46,920 --> 00:57:50,960
is that the time it takes to generate is

1389
00:57:49,200 --> 00:57:52,640
proportional to the amount of points or

1390
00:57:50,960 --> 00:57:54,839
the product of the length and the height

1391
00:57:52,640 --> 00:57:56,839
multiplied by themselves so in Big O

1392
00:57:54,839 --> 00:58:00,960
notation that's o of n s

1393
00:57:56,839 --> 00:58:02,680
and we do not like quadratic uh Bigos no

1394
00:58:00,960 --> 00:58:04,760
um so it kind of goes without saying I

1395
00:58:02,680 --> 00:58:07,039
knew I'd have to optimize it eventually

1396
00:58:04,760 --> 00:58:09,720
but to start with I ignored it because I

1397
00:58:07,039 --> 00:58:12,520
didn't know how um I was still getting

1398
00:58:09,720 --> 00:58:14,400
the hang of python at this point and so

1399
00:58:12,520 --> 00:58:16,559
I decided to get it working before I

1400
00:58:14,400 --> 00:58:19,280
tried to fix this uh because you can't

1401
00:58:16,559 --> 00:58:22,799
like make something better if it's not

1402
00:58:19,280 --> 00:58:24,680
there at all um but I had a fix in my

1403
00:58:22,799 --> 00:58:26,720
mind but I had zero clue on how to

1404
00:58:24,680 --> 00:58:28,799
implement it

1405
00:58:26,720 --> 00:58:31,240
uh so this is the first uh sort of

1406
00:58:28,799 --> 00:58:33,280
iteration of the code uh start by

1407
00:58:31,240 --> 00:58:35,160
importing some libraries and defining

1408
00:58:33,280 --> 00:58:37,359
some parameters so we've got starting

1409
00:58:35,160 --> 00:58:39,000
complexity which is the number of random

1410
00:58:37,359 --> 00:58:41,319
points we start off with with completely

1411
00:58:39,000 --> 00:58:43,319
random values uh we've got the

1412
00:58:41,319 --> 00:58:45,240
iterations total number of points to be

1413
00:58:43,319 --> 00:58:47,000
placed because for this one for some

1414
00:58:45,240 --> 00:58:48,839
reason I wasn't doing the shuffling the

1415
00:58:47,000 --> 00:58:50,400
random coordinates I was just placing

1416
00:58:48,839 --> 00:58:52,799
them at random places and hoping it

1417
00:58:50,400 --> 00:58:55,039
worked out uh and then the resolution

1418
00:58:52,799 --> 00:58:56,559
which is the size of the place pixels so

1419
00:58:55,039 --> 00:58:59,039
you can increase the res res olution to

1420
00:58:56,559 --> 00:59:02,119
make it sort of blob and decrease it to

1421
00:58:59,039 --> 00:59:05,880
make it uh not hide the bits it missed

1422
00:59:02,119 --> 00:59:07,640
as much um so the yeah instead of

1423
00:59:05,880 --> 00:59:09,880
randomizing the coordinate list points

1424
00:59:07,640 --> 00:59:11,559
are generated with random coordinates

1425
00:59:09,880 --> 00:59:14,160
and yeah the main problems with that

1426
00:59:11,559 --> 00:59:15,960
approach were that uh one you couldn't

1427
00:59:14,160 --> 00:59:17,799
guarantee that points wouldn't be

1428
00:59:15,960 --> 00:59:18,960
repeated it wouldn't overlap one point

1429
00:59:17,799 --> 00:59:21,520
with another one that had already been

1430
00:59:18,960 --> 00:59:23,920
done and two you couldn't guarantee that

1431
00:59:21,520 --> 00:59:26,960
every point was hit um not sure why I

1432
00:59:23,920 --> 00:59:29,839
did it like that but I did

1433
00:59:26,960 --> 00:59:31,880
uh so yeah changing the resolution can

1434
00:59:29,839 --> 00:59:34,720
compensate for any missed or overlapped

1435
00:59:31,880 --> 00:59:36,599
parts and a final step called Smooth

1436
00:59:34,720 --> 00:59:38,079
iterations involves processing every

1437
00:59:36,599 --> 00:59:40,720
point to set it to the average of its

1438
00:59:38,079 --> 00:59:44,119
neighbors uh average of its nonzero

1439
00:59:40,720 --> 00:59:45,680
neighbors um and you average it onto a

1440
00:59:44,119 --> 00:59:47,000
buffer frame so that you're not

1441
00:59:45,680 --> 00:59:48,240
averaging ones that have already been

1442
00:59:47,000 --> 00:59:49,000
averaged because that can create like

1443
00:59:48,240 --> 00:59:51,880
weird

1444
00:59:49,000 --> 00:59:53,559
smears create Point function uh it will

1445
00:59:51,880 --> 00:59:55,799
generate a random X and Y position

1446
00:59:53,559 --> 00:59:58,680
within the bounds of the screen called

1447
00:59:55,799 --> 01:00:00,720
uh oh GL get closest points and set the

1448
00:59:58,680 --> 01:00:03,520
height to the average and add that to

1449
01:00:00,720 --> 01:00:05,400
the list add that point to the list get

1450
01:00:03,520 --> 01:00:08,200
closest points is such a great function

1451
01:00:05,400 --> 01:00:09,799
that I have no clue how it works um I

1452
01:00:08,200 --> 01:00:11,680
yed it from like stack Overflow or

1453
01:00:09,799 --> 01:00:14,000
something when I was first making this

1454
01:00:11,680 --> 01:00:16,079
because I was using Pythagoras Theorem

1455
01:00:14,000 --> 01:00:18,039
to just find the closest points but that

1456
01:00:16,079 --> 01:00:20,079
was really slow and I wanted to speed it

1457
01:00:18,039 --> 01:00:22,880
up it still took like 6 hours to

1458
01:00:20,079 --> 01:00:24,799
generate like a 1,00 by 1000 pixel thing

1459
01:00:22,880 --> 01:00:27,400
though um so yeah originally used

1460
01:00:24,799 --> 01:00:30,440
Pythagoras theem to find the

1461
01:00:27,400 --> 01:00:32,799
points uh and then for Loops we create

1462
01:00:30,440 --> 01:00:33,760
all the starting points create all the

1463
01:00:32,799 --> 01:00:37,079
point

1464
01:00:33,760 --> 01:00:38,880
points and in the main Loop we have some

1465
01:00:37,079 --> 01:00:41,599
basic uh the main Loop is just one in

1466
01:00:38,880 --> 01:00:43,480
the like bottom rsh sort of corner we

1467
01:00:41,599 --> 01:00:45,119
have some basic P game Necessities cuz

1468
01:00:43,480 --> 01:00:47,000
Pi gamees the library I was using to

1469
01:00:45,119 --> 01:00:49,680
visualize all this so I could update the

1470
01:00:47,000 --> 01:00:52,559
C Level because I am visualizing the

1471
01:00:49,680 --> 01:00:54,400
noise in this one as sort of like red I

1472
01:00:52,559 --> 01:00:56,440
mean blue and green

1473
01:00:54,400 --> 01:00:59,960
Islands um so yeah we Loop over

1474
01:00:56,440 --> 01:01:01,400
everything Set uh each Point's color to

1475
01:00:59,960 --> 01:01:03,000
a function of the sea level and its

1476
01:01:01,400 --> 01:01:05,400
height and then draw a circle of radius

1477
01:01:03,000 --> 01:01:07,280
equal to the resolution I was pretty

1478
01:01:05,400 --> 01:01:08,920
happy with that um I had to leave it

1479
01:01:07,280 --> 01:01:10,280
running in the morning go to school and

1480
01:01:08,920 --> 01:01:13,319
come back and it still wouldn't be

1481
01:01:10,280 --> 01:01:15,160
finished but uh it still worked and that

1482
01:01:13,319 --> 01:01:18,200
was good enough for

1483
01:01:15,160 --> 01:01:20,520
now um eventually though at the start of

1484
01:01:18,200 --> 01:01:22,799
2024 I was working on like a really

1485
01:01:20,520 --> 01:01:24,319
boring assignment for English and I was

1486
01:01:22,799 --> 01:01:26,640
desperate to procrastinate so I opened

1487
01:01:24,319 --> 01:01:29,039
up my python IDE and look through my old

1488
01:01:26,640 --> 01:01:30,480
projects before finding this one again I

1489
01:01:29,039 --> 01:01:32,599
decided I was ready to tackle the

1490
01:01:30,480 --> 01:01:34,760
problem of optimization and my idea for

1491
01:01:32,599 --> 01:01:36,960
this was to use

1492
01:01:34,760 --> 01:01:38,280
trunks at the moment due to the fact

1493
01:01:36,960 --> 01:01:40,280
that I was drawing all the points in a

1494
01:01:38,280 --> 01:01:41,720
single list to find any information

1495
01:01:40,280 --> 01:01:43,520
about the distances to the reference

1496
01:01:41,720 --> 01:01:45,680
point I had to Loop over every single

1497
01:01:43,520 --> 01:01:46,960
point when as a human I'd be able to say

1498
01:01:45,680 --> 01:01:48,280
I don't need to worry about those points

1499
01:01:46,960 --> 01:01:51,599
or those points I can just worry about

1500
01:01:48,280 --> 01:01:53,640
the ones right near it to fix this uh I

1501
01:01:51,599 --> 01:01:56,279
can store each point in a chunk so as

1502
01:01:53,640 --> 01:01:58,319
I've shown there each chunk uh led by

1503
01:01:56,279 --> 01:02:00,319
the S like Boulder lines represents a

1504
01:01:58,319 --> 01:02:02,640
portion of the screen that's X pixels by

1505
01:02:00,319 --> 01:02:03,920
X pixels big and these trunks would be

1506
01:02:02,640 --> 01:02:06,440
arranged in a grid

1507
01:02:03,920 --> 01:02:08,720
pattern I can then access the trunk of

1508
01:02:06,440 --> 01:02:10,400
any specific point shown by the Red Dot

1509
01:02:08,720 --> 01:02:12,520
and only check its trunk and the

1510
01:02:10,400 --> 01:02:13,799
neighboring trunks for close points this

1511
01:02:12,520 --> 01:02:16,079
means that the total amount of points

1512
01:02:13,799 --> 01:02:18,240
I'm looping over for any one point goes

1513
01:02:16,079 --> 01:02:21,799
from n where n is the amount of pixels

1514
01:02:18,240 --> 01:02:23,920
on the sample down to 3 * the chunk size

1515
01:02:21,799 --> 01:02:26,319
squared where C is the size of each

1516
01:02:23,920 --> 01:02:28,480
chunk and because C is a constant that

1517
01:02:26,319 --> 01:02:31,520
doesn't change with the input uh pixel

1518
01:02:28,480 --> 01:02:33,240
um mounts uh this is now a linear

1519
01:02:31,520 --> 01:02:36,760
scaling which is a lot better than

1520
01:02:33,240 --> 01:02:38,839
quadratic yay uh but to implement this I

1521
01:02:36,760 --> 01:02:40,520
needed to create a 2d Matrix of chunks

1522
01:02:38,839 --> 01:02:42,640
where each chunk contained in of itself

1523
01:02:40,520 --> 01:02:45,480
a 2d Matrix of points making for a 40

1524
01:02:42,640 --> 01:02:47,279
Matrix in total um and then I can just

1525
01:02:45,480 --> 01:02:51,240
reference a specific point in The Matrix

1526
01:02:47,279 --> 01:02:53,599
with like Matrix trunk X trunk y point x

1527
01:02:51,240 --> 01:02:54,960
point Y and that's actually it's like

1528
01:02:53,599 --> 01:02:57,160
python is so nice to that sort of thing

1529
01:02:54,960 --> 01:02:59,480
it's so easy

1530
01:02:57,160 --> 01:03:02,480
uh so it starts off the same uh reset

1531
01:02:59,480 --> 01:03:06,559
chunk map just generates a 40 Matrix

1532
01:03:02,480 --> 01:03:09,240
with every Point height set to ne1 we uh

1533
01:03:06,559 --> 01:03:11,799
generate some complexity points more for

1534
01:03:09,240 --> 01:03:14,640
Loops uh check if the random value

1535
01:03:11,799 --> 01:03:16,559
between zero and point Chan is one if so

1536
01:03:14,640 --> 01:03:18,839
it sets that point to a random height so

1537
01:03:16,559 --> 01:03:21,000
so for this version we are starting off

1538
01:03:18,839 --> 01:03:22,799
by putting random uh putting random

1539
01:03:21,000 --> 01:03:25,200
points with a random chance in each

1540
01:03:22,799 --> 01:03:27,160
trunk instead of putting a predetermined

1541
01:03:25,200 --> 01:03:28,240
amount of trunks uh points randomly in

1542
01:03:27,160 --> 01:03:31,400
every

1543
01:03:28,240 --> 01:03:33,920
chunk um and then everything oh I love

1544
01:03:31,400 --> 01:03:34,760
all Chords it's like some monstrosity of

1545
01:03:33,920 --> 01:03:38,559
list

1546
01:03:34,760 --> 01:03:40,039
comprehension um it hurts me to look at

1547
01:03:38,559 --> 01:03:41,319
but I think I probably learned about it

1548
01:03:40,039 --> 01:03:46,319
a couple weeks before and I was like

1549
01:03:41,319 --> 01:03:49,920
this is a great use case um it wasn't

1550
01:03:46,319 --> 01:03:51,119
so yeah uh generate points Loops over

1551
01:03:49,920 --> 01:03:53,240
everything this is all just kind of

1552
01:03:51,119 --> 01:03:56,039
boring generate points is a kind of

1553
01:03:53,240 --> 01:03:58,760
interesting function we uh have d X and

1554
01:03:56,039 --> 01:04:00,200
Dy and they're uh sort of like offsets

1555
01:03:58,760 --> 01:04:02,200
we Loop over them check if they're in

1556
01:04:00,200 --> 01:04:05,960
bounds and then we do some funky maths

1557
01:04:02,200 --> 01:04:07,240
to find the points and yeah um but now

1558
01:04:05,960 --> 01:04:10,359
I'm running out of times so I really

1559
01:04:07,240 --> 01:04:11,119
want to quickly showcase some of the

1560
01:04:10,359 --> 01:04:15,400
cool

1561
01:04:11,119 --> 01:04:17,599
patterns uh this is a 300X 400 one and

1562
01:04:15,400 --> 01:04:19,359
this is a fractal uh iterative Point

1563
01:04:17,599 --> 01:04:21,720
averaging one and it's created by

1564
01:04:19,359 --> 01:04:24,160
getting a large one which uses large

1565
01:04:21,720 --> 01:04:26,720
trunk sizes and a low amount of points

1566
01:04:24,160 --> 01:04:30,240
per trunk and cre creating a small one

1567
01:04:26,720 --> 01:04:34,960
and then overlapping them to create

1568
01:04:30,240 --> 01:04:37,920
this I also made this little uh one uh

1569
01:04:34,960 --> 01:04:39,279
file which will do it all and it will do

1570
01:04:37,920 --> 01:04:42,559
it live but it's showing on the wrong

1571
01:04:39,279 --> 01:04:46,000
display so really quickly no okay well I

1572
01:04:42,559 --> 01:04:48,680
think my time's out so thank you oh no

1573
01:04:46,000 --> 01:04:51,200
go on go on put the

1574
01:04:48,680 --> 01:04:53,119
display get Give It 2 seconds do you

1575
01:04:51,200 --> 01:04:55,960
have the display Give It 2 seconds go on

1576
01:04:53,119 --> 01:04:58,680
go on go on you can't just leave was

1577
01:04:55,960 --> 01:04:58,680
hanging like

1578
01:04:58,880 --> 01:05:08,200
that yeah I can't wait that long your

1579
01:05:03,720 --> 01:05:11,760
Cod and make fix your code

1580
01:05:08,200 --> 01:05:15,680
live love it how hard how hard could

1581
01:05:11,760 --> 01:05:19,079
that be how hard oh ER is Windows P yeah

1582
01:05:15,680 --> 01:05:20,920
I just got a Windows p Ah but the thing

1583
01:05:19,079 --> 01:05:24,599
is it's a collaborative and then

1584
01:05:20,920 --> 01:05:26,599
duplicate debugging session see

1585
01:05:24,599 --> 01:05:29,839
fantastic okay

1586
01:05:26,599 --> 01:05:31,279
pair programming with all of everyone

1587
01:05:29,839 --> 01:05:33,839
now look we get to watch it Place its

1588
01:05:31,279 --> 01:05:33,839
points as it

1589
01:05:37,079 --> 01:05:41,599
goes so

1590
01:05:39,240 --> 01:05:44,799
pretty and I mean it takes a minute but

1591
01:05:41,599 --> 01:05:47,799
I'm like a 100x 100 one in the updated

1592
01:05:44,799 --> 01:05:50,440
version takes uh .9 seconds on average

1593
01:05:47,799 --> 01:05:53,480
whereas in the old version with n s

1594
01:05:50,440 --> 01:05:56,079
complexity it took 70 second uh 77

1595
01:05:53,480 --> 01:05:58,000
seconds which wasn't great

1596
01:05:56,079 --> 01:05:59,920
uh but yeah we'll keep doing that if

1597
01:05:58,000 --> 01:06:03,599
once it's all filled in it will color it

1598
01:05:59,920 --> 01:06:03,599
blue and green depending on the sea

1599
01:06:07,400 --> 01:06:11,880
level I think this is the most Zen I've

1600
01:06:10,160 --> 01:06:14,640
felt in

1601
01:06:11,880 --> 01:06:17,520
months we

1602
01:06:14,640 --> 01:06:20,000
just yeah it's a crackling fire I'm

1603
01:06:17,520 --> 01:06:23,440
getting crackling fire Vibes the new

1604
01:06:20,000 --> 01:06:26,839
screen for everybody yes this is

1605
01:06:23,440 --> 01:06:29,279
definitely the new screen saver yeah

1606
01:06:26,839 --> 01:06:30,720
all right everyone can you take a big

1607
01:06:29,279 --> 01:06:34,520
breath

1608
01:06:30,720 --> 01:06:34,520
in and a big breath

1609
01:06:35,000 --> 01:06:40,200
out I think we felt some Zen yeah

1610
01:06:38,240 --> 01:06:41,599
there's also some like errors around the

1611
01:06:40,200 --> 01:06:43,160
sides with this version I'm not sure

1612
01:06:41,599 --> 01:06:46,000
where they came from but I just want to

1613
01:06:43,160 --> 01:06:47,400
point them out Zen focus on the Zen

1614
01:06:46,000 --> 01:06:50,880
there are no mistakes only happy

1615
01:06:47,400 --> 01:06:54,160
accidents happy accidents indeed can you

1616
01:06:50,880 --> 01:06:57,160
all give Toby a massive Round of

1617
01:06:54,160 --> 01:06:57,160
Applause

1618
01:07:03,200 --> 01:07:08,599
there is one more

1619
01:07:05,910 --> 01:07:12,559
[Music]

1620
01:07:08,599 --> 01:07:20,359
talk there is one more talk and while

1621
01:07:12,559 --> 01:07:20,359
they get set up oh I will intro them and

1622
01:07:20,559 --> 01:07:26,599
my the fear in my voice is not not

1623
01:07:24,520 --> 01:07:30,039
actually because of the potential fire

1624
01:07:26,599 --> 01:07:32,400
risk cool um but actually because now I

1625
01:07:30,039 --> 01:07:35,000
have three separate emoji that I need to

1626
01:07:32,400 --> 01:07:39,400
make you all

1627
01:07:35,000 --> 01:07:43,559
guess no none of them are fire Emoji um

1628
01:07:39,400 --> 01:07:47,680
okay so hang on while we do this now we

1629
01:07:43,559 --> 01:07:52,760
will be joined by Derek wave good don't

1630
01:07:47,680 --> 01:07:55,960
drop the robot um helius hi wave and

1631
01:07:52,760 --> 01:08:00,000
Matthew wave um all right d has been

1632
01:07:55,960 --> 01:08:04,480
programming for 4 years um his favorite

1633
01:08:00,000 --> 01:08:04,480
emoji is

1634
01:08:04,720 --> 01:08:13,720
um yeah shark Emoji excellent uh and he

1635
01:08:09,279 --> 01:08:15,680
has been learning Cur uh reinforcement

1636
01:08:13,720 --> 01:08:16,920
learning for the past year and recently

1637
01:08:15,680 --> 01:08:19,799
discovered how it can be applied to

1638
01:08:16,920 --> 01:08:21,600
train agents to make decisions in real

1639
01:08:19,799 --> 01:08:23,120
life environments it's fascinating to

1640
01:08:21,600 --> 01:08:25,159
see how algorithms can learn through

1641
01:08:23,120 --> 01:08:26,000
trial and error much like humans

1642
01:08:25,159 --> 01:08:29,920
actually

1643
01:08:26,000 --> 01:08:31,400
um yes that is great okay uh hel oh gosh

1644
01:08:29,920 --> 01:08:33,719
helus has been programming for seven

1645
01:08:31,400 --> 01:08:35,679
years heli's favorite emoji oh wait

1646
01:08:33,719 --> 01:08:37,960
you've got you've

1647
01:08:35,679 --> 01:08:40,759
moved I feel like we're having a weird

1648
01:08:37,960 --> 01:08:40,759
mime thing

1649
01:08:41,759 --> 01:08:46,040
now sorry they were I I heard things but

1650
01:08:44,359 --> 01:08:48,319
I didn't hear what they

1651
01:08:46,040 --> 01:08:52,799
were

1652
01:08:48,319 --> 01:08:55,719
uh it's a little bit more green and red

1653
01:08:52,799 --> 01:09:00,159
watermelon yes

1654
01:08:55,719 --> 01:09:04,799
nice nice um and hel is like that most

1655
01:09:00,159 --> 01:09:07,920
webcams can capture infrared light o

1656
01:09:04,799 --> 01:09:09,640
file that away under useful maybe and

1657
01:09:07,920 --> 01:09:12,199
Matthew has been programming for seven

1658
01:09:09,640 --> 01:09:14,279
years as well I asked Matthew about this

1659
01:09:12,199 --> 01:09:17,520
Emoji Choice specifically asked him what

1660
01:09:14,279 --> 01:09:20,159
it was called he said he didn't know he

1661
01:09:17,520 --> 01:09:24,040
just chose it randomly do do you want me

1662
01:09:20,159 --> 01:09:26,480
to choose a different one um so I don't

1663
01:09:24,040 --> 01:09:31,239
I I mean the

1664
01:09:26,480 --> 01:09:34,719
th this is the Emoji the Pinchy

1665
01:09:31,239 --> 01:09:37,159
fingers I yeah I'm this close Okay I get

1666
01:09:34,719 --> 01:09:41,799
okay I'm this close we're this we're

1667
01:09:37,159 --> 01:09:44,319
this close to to this close this big um

1668
01:09:41,799 --> 01:09:45,359
and Matthew recently M wanted to share

1669
01:09:44,319 --> 01:09:47,880
something that he recently learned which

1670
01:09:45,359 --> 01:09:49,359
is the only Windows uh uh the only

1671
01:09:47,880 --> 01:09:52,080
reason that Windows is the standard is

1672
01:09:49,359 --> 01:09:57,080
due to marketing

1673
01:09:52,080 --> 01:09:58,280
um yeah um those those were our Emoji

1674
01:09:57,080 --> 01:10:01,920
thank you for sharing them those were

1675
01:09:58,280 --> 01:10:04,840
our facts thank you for sharing them um

1676
01:10:01,920 --> 01:10:10,480
I get to keep stalling

1677
01:10:04,840 --> 01:10:12,480
until uh we get there um I would but I

1678
01:10:10,480 --> 01:10:13,560
there's something in my like in the

1679
01:10:12,480 --> 01:10:15,800
dance

1680
01:10:13,560 --> 01:10:18,320
area dance down there but it wouldn't be

1681
01:10:15,800 --> 01:10:20,480
captured on the film and it's just it

1682
01:10:18,320 --> 01:10:22,400
would be unfair if I didn't get to share

1683
01:10:20,480 --> 01:10:24,199
you need the tape no good okay maybe I

1684
01:10:22,400 --> 01:10:26,920
will dance down

1685
01:10:24,199 --> 01:10:30,000
here um um maybe we all need just need

1686
01:10:26,920 --> 01:10:33,120
to breathe again and keep Imagining the

1687
01:10:30,000 --> 01:10:38,360
uh the Zen noise we have

1688
01:10:33,120 --> 01:10:41,080
a what's my favorite emoji actually my

1689
01:10:38,360 --> 01:10:45,800
this this sounds like I'm just pandering

1690
01:10:41,080 --> 01:10:45,800
now but my favorite emoji is the parrot

1691
01:10:46,000 --> 01:10:53,440
Emoji because I oh I I just spoiled it

1692
01:10:49,520 --> 01:10:53,440
I'm so sorry I'm like

1693
01:10:54,460 --> 01:10:59,480
[Applause]

1694
01:10:57,400 --> 01:11:03,120
yeah all all the way the through the

1695
01:10:59,480 --> 01:11:03,120
parro Emoji

1696
01:11:04,719 --> 01:11:10,640
um yeah well since we find ourselves

1697
01:11:08,520 --> 01:11:12,600
here in Melbourne and I don't often come

1698
01:11:10,640 --> 01:11:14,440
down to Melbourne maybe I should just

1699
01:11:12,600 --> 01:11:16,719
make a pitch for the girls Programming

1700
01:11:14,440 --> 01:11:19,400
Network which has a chapter here in

1701
01:11:16,719 --> 01:11:23,199
Melbourne and and there are free

1702
01:11:19,400 --> 01:11:25,440
workshops for high school girls to uh

1703
01:11:23,199 --> 01:11:29,040
and gender diverse kids to learn

1704
01:11:25,440 --> 01:11:32,560
programming and have fun and get pizza

1705
01:11:29,040 --> 01:11:35,239
cuz you know yeah yeah come come for the

1706
01:11:32,560 --> 01:11:38,080
pizza stay for the friends um but if

1707
01:11:35,239 --> 01:11:40,280
you're not in Melbourne maybe you're in

1708
01:11:38,080 --> 01:11:42,920
Sydney guess what there's there's a

1709
01:11:40,280 --> 01:11:44,920
chapter for you there too maybe you're

1710
01:11:42,920 --> 01:11:48,639
actually from Perth hey there's a

1711
01:11:44,920 --> 01:11:51,120
chapter there too um canra oh yeah we

1712
01:11:48,639 --> 01:11:53,679
got you covered don't worry com next

1713
01:11:51,120 --> 01:11:56,600
year two more coming next year so many

1714
01:11:53,679 --> 01:11:58,880
so many things Adelaide and

1715
01:11:56,600 --> 01:12:02,560
Brisbane um but if you don't live in any

1716
01:11:58,880 --> 01:12:05,639
of those places and you want a gpn

1717
01:12:02,560 --> 01:12:08,360
network near you and you want to give me

1718
01:12:05,639 --> 01:12:08,360
and Renee more

1719
01:12:10,080 --> 01:12:15,280
work yeah uh and yourself more work as

1720
01:12:13,360 --> 01:12:17,880
well yeah hit us up um does anyone else

1721
01:12:15,280 --> 01:12:22,639
want anything spooked since I'm

1722
01:12:17,880 --> 01:12:25,760
here oh yes we we also need uh fabulous

1723
01:12:22,639 --> 01:12:27,840
volunteers all of the time wonderful um

1724
01:12:25,760 --> 01:12:32,920
have we all been enjoying

1725
01:12:27,840 --> 01:12:35,320
Pon yes did you know you could volunteer

1726
01:12:32,920 --> 01:12:37,400
at Pon as

1727
01:12:35,320 --> 01:12:41,120
well

1728
01:12:37,400 --> 01:12:44,880
um um okay how are we

1729
01:12:41,120 --> 01:12:48,239
going are we good we're good we're good

1730
01:12:44,880 --> 01:12:49,320
maybe rotate the thing away from the CR

1731
01:12:48,239 --> 01:12:52,080
what

1732
01:12:49,320 --> 01:12:56,239
no what do you mean rotate it away from

1733
01:12:52,080 --> 01:12:56,239
the crowd give a fight

1734
01:12:57,440 --> 01:13:03,480
did you just aim it at yourselves

1735
01:13:00,120 --> 01:13:05,360
instead of at the crowd yeah I mean

1736
01:13:03,480 --> 01:13:08,800
replaceable

1737
01:13:05,360 --> 01:13:10,440
yeah I feel like I feel like I'm in I

1738
01:13:08,800 --> 01:13:12,280
feel like I'm at the zoo at the seal

1739
01:13:10,440 --> 01:13:14,480
show and I'm need to warn the crowd that

1740
01:13:12,280 --> 01:13:16,840
the first two rows are the Splash

1741
01:13:14,480 --> 01:13:19,040
Zone but actually is this the Splash

1742
01:13:16,840 --> 01:13:21,560
Zone I I feel like we okay hang on let

1743
01:13:19,040 --> 01:13:23,120
me let me see a view of hands this area

1744
01:13:21,560 --> 01:13:24,120
of the crowd do you want to be the

1745
01:13:23,120 --> 01:13:27,120
Splash

1746
01:13:24,120 --> 01:13:30,280
Zone not if it's fire How likely is it

1747
01:13:27,120 --> 01:13:32,360
going to be fire or confetti I think it

1748
01:13:30,280 --> 01:13:35,000
should be all right I I think we're

1749
01:13:32,360 --> 01:13:36,600
happy to be the Splash Zone yeah come on

1750
01:13:35,000 --> 01:13:39,960
live dangerously let us be the Splash

1751
01:13:36,600 --> 01:13:41,520
Zone okay a aim it towards the camera

1752
01:13:39,960 --> 01:13:44,600
that's right think of the poor people

1753
01:13:41,520 --> 01:13:51,080
watching this later on they want to be

1754
01:13:44,600 --> 01:13:53,600
involved okay I'm not here with o um so

1755
01:13:51,080 --> 01:13:57,040
can I yeah yeah yeah what's duty of care

1756
01:13:53,600 --> 01:13:57,040
I'm not that teacher okay

1757
01:13:57,880 --> 01:14:03,800
so can we welcome last talk up take it

1758
01:14:07,280 --> 01:14:13,080
away hi um we're presenting Bev it is a

1759
01:14:10,400 --> 01:14:14,880
bip bipedal walking confetti ejecting

1760
01:14:13,080 --> 01:14:16,800
robot with voice control it's

1761
01:14:14,880 --> 01:14:18,120
essentially a confetti launcher mounted

1762
01:14:16,800 --> 01:14:19,159
on a pair of legs that follow you around

1763
01:14:18,120 --> 01:14:21,600
and you can talk to it and give it

1764
01:14:19,159 --> 01:14:25,560
commands with voice controls uh so who

1765
01:14:21,600 --> 01:14:27,920
are we I'm Matthew on my next here is

1766
01:14:25,560 --> 01:14:30,239
Derek and then Helios on the very edge

1767
01:14:27,920 --> 01:14:34,040
um Derek has been doing python C++ and

1768
01:14:30,239 --> 01:14:37,719
Lua um he's broken a couple things um so

1769
01:14:34,040 --> 01:14:39,840
was so of all of us really but uh

1770
01:14:37,719 --> 01:14:41,880
yeah um so this is what we've done right

1771
01:14:39,840 --> 01:14:43,560
so Derek was the main legs component

1772
01:14:41,880 --> 01:14:44,800
Helios did the main confetti dispenser

1773
01:14:43,560 --> 01:14:47,400
and that was the voice controls that

1774
01:14:44,800 --> 01:14:50,639
kind of put them together uh first up

1775
01:14:47,400 --> 01:14:50,639
I'll give it to Derek to talk about his

1776
01:14:52,120 --> 01:14:58,080
legs okay so um yeah I made the bipedal

1777
01:14:55,480 --> 01:15:00,080
walking uh robot um using reinforcement

1778
01:14:58,080 --> 01:15:02,400
learning uh if you don't know what

1779
01:15:00,080 --> 01:15:05,960
reinforcement learning is it's basically

1780
01:15:02,400 --> 01:15:07,800
just um a discipline where okay thank

1781
01:15:05,960 --> 01:15:10,239
you it's basically just a discipline

1782
01:15:07,800 --> 01:15:13,159
where it uses uh trial and error to

1783
01:15:10,239 --> 01:15:16,159
figure out like I know why it exists and

1784
01:15:13,159 --> 01:15:17,560
basically like it's like it basically

1785
01:15:16,159 --> 01:15:22,000
tries to figure out what its goal is and

1786
01:15:17,560 --> 01:15:24,520
try to maximize what its goal um wait

1787
01:15:22,000 --> 01:15:26,880
okay to better explain this we can use a

1788
01:15:24,520 --> 01:15:29,480
like a simple example of a dog in an

1789
01:15:26,880 --> 01:15:32,280
entree dog in a yard if we like throw a

1790
01:15:29,480 --> 01:15:34,920
stick the dog will first just like

1791
01:15:32,280 --> 01:15:36,639
explore around for a bit and but

1792
01:15:34,920 --> 01:15:39,159
eventually you'll find the stick and

1793
01:15:36,639 --> 01:15:41,000
bring it back to us um when it brings it

1794
01:15:39,159 --> 01:15:43,280
back to us uh we'll give it a treat or a

1795
01:15:41,000 --> 01:15:45,639
reward and then the next time we throw

1796
01:15:43,280 --> 01:15:47,320
the stick around um the dog won't

1797
01:15:45,639 --> 01:15:49,480
explore around as much but go straight

1798
01:15:47,320 --> 01:15:52,600
to the stick and bring it back to us

1799
01:15:49,480 --> 01:15:55,360
that is the basics of reinforcement Lear

1800
01:15:52,600 --> 01:15:57,120
and um the next thing I'll talk about is

1801
01:15:55,360 --> 01:16:01,440
the the type of reinforcement learning

1802
01:15:57,120 --> 01:16:04,159
algorithm that I chose so I chose

1803
01:16:01,440 --> 01:16:06,679
td3 uh which stands for twin delayed

1804
01:16:04,159 --> 01:16:10,600
deep deterministic policy gradient yeah

1805
01:16:06,679 --> 01:16:12,320
it's a mouthful but uh td3 is basically

1806
01:16:10,600 --> 01:16:15,120
just an algorithm that helps the

1807
01:16:12,320 --> 01:16:19,320
computer program to make um good

1808
01:16:15,120 --> 01:16:22,000
decisions um the T3 has two components

1809
01:16:19,320 --> 01:16:23,920
uh one is the actor uh it's basically

1810
01:16:22,000 --> 01:16:25,639
just like a coach that decides what

1811
01:16:23,920 --> 01:16:28,080
moves to make in a game

1812
01:16:25,639 --> 01:16:30,360
um like it looks at the game and says

1813
01:16:28,080 --> 01:16:32,600
let's try this move and the critic is

1814
01:16:30,360 --> 01:16:36,360
basically just a judge that says that

1815
01:16:32,600 --> 01:16:39,760
was either a good move or was a bad

1816
01:16:36,360 --> 01:16:42,800
move um what's special about td3 it has

1817
01:16:39,760 --> 01:16:44,639
two critics um this is because sometimes

1818
01:16:42,800 --> 01:16:48,679
one judge might be a little too

1819
01:16:44,639 --> 01:16:51,560
optimistic and say a bad move uh is good

1820
01:16:48,679 --> 01:16:53,880
is good but by having two judges the

1821
01:16:51,560 --> 01:16:56,679
agent only listens to the more cautious

1822
01:16:53,880 --> 01:16:58,800
one or safe one and this makes learning

1823
01:16:56,679 --> 01:17:01,320
more steady and

1824
01:16:58,800 --> 01:17:06,440
reliable another thing it utilizes is

1825
01:17:01,320 --> 01:17:11,880
delayed learning so um uh in a game what

1826
01:17:06,440 --> 01:17:15,360
was I saying uh yeah okay um in delay it

1827
01:17:11,880 --> 01:17:18,679
uses delayed learning where um in a

1828
01:17:15,360 --> 01:17:20,679
game it okay I forget forgot what I was

1829
01:17:18,679 --> 01:17:22,080
going to say but it basically just lets

1830
01:17:20,679 --> 01:17:24,920
the actor or the coach change its

1831
01:17:22,080 --> 01:17:26,679
strategy only after a few moves um this

1832
01:17:24,920 --> 01:17:28,960
gives it uh time to learn from better

1833
01:17:26,679 --> 01:17:31,520
feedback uh better feedback from the

1834
01:17:28,960 --> 01:17:35,719
critic uh the next thing it utilizes is

1835
01:17:31,520 --> 01:17:37,920
noise so in a game um uh to better

1836
01:17:35,719 --> 01:17:43,120
explain this we can use a game example

1837
01:17:37,920 --> 01:17:44,960
where um the U the the robot would use

1838
01:17:43,120 --> 01:17:47,400
the exact same strategy every time in

1839
01:17:44,960 --> 01:17:51,199
the game but this is bad because it will

1840
01:17:47,400 --> 01:17:53,000
it reduces exploration and this um stops

1841
01:17:51,199 --> 01:17:56,600
it from learning new strategies or

1842
01:17:53,000 --> 01:17:58,639
better strategies in the long run um td3

1843
01:17:56,600 --> 01:18:00,719
um improves this by adding a little

1844
01:17:58,639 --> 01:18:03,159
noise or Randomness to the moves like

1845
01:18:00,719 --> 01:18:04,480
trying something slightly different to

1846
01:18:03,159 --> 01:18:07,159
discover better

1847
01:18:04,480 --> 01:18:09,440
options and that's a really basic

1848
01:18:07,159 --> 01:18:12,280
rundown of the algorithm now onto the

1849
01:18:09,440 --> 01:18:14,760
build of the robot the robot is built on

1850
01:18:12,280 --> 01:18:17,360
a 3D printer frame where it is connected

1851
01:18:14,760 --> 01:18:20,040
to 10 servos three arduinos and a

1852
01:18:17,360 --> 01:18:22,400
Raspberry Pi uh there's also a camera on

1853
01:18:20,040 --> 01:18:24,920
board um which is again used for

1854
01:18:22,400 --> 01:18:26,920
computer vision um the entire robot is

1855
01:18:24,920 --> 01:18:32,679
powered by an ATX power supply that I

1856
01:18:26,920 --> 01:18:36,880
found at the dump but yeah but yeah

1857
01:18:32,679 --> 01:18:40,520
um uh the the sensors that I used for

1858
01:18:36,880 --> 01:18:44,880
the robot um um well actually let me

1859
01:18:40,520 --> 01:18:46,719
give a example um the senses uh are used

1860
01:18:44,880 --> 01:18:49,320
in the robot so it can learn from its

1861
01:18:46,719 --> 01:18:51,480
environment so it can optimize its goal

1862
01:18:49,320 --> 01:18:53,840
uh the sensors that I used were an MP

1863
01:18:51,480 --> 01:18:55,560
6050 which gives the robot its

1864
01:18:53,840 --> 01:18:57,600
orientation data

1865
01:18:55,560 --> 01:19:00,040
um there was acceleration data but I

1866
01:18:57,600 --> 01:19:03,800
didn't use it for some reason I don't

1867
01:19:00,040 --> 01:19:06,639
know um the next uh the next sensor I

1868
01:19:03,800 --> 01:19:09,960
used was a camera um this uh gives the

1869
01:19:06,639 --> 01:19:12,600
robot data on how far a human is this is

1870
01:19:09,960 --> 01:19:15,800
relevant to his project CU he said oh I

1871
01:19:12,600 --> 01:19:18,320
need the confetti close to a person so

1872
01:19:15,800 --> 01:19:21,679
yeah so I was like okay sure I'll give a

1873
01:19:18,320 --> 01:19:23,840
camera to it um the next uh sensor I

1874
01:19:21,679 --> 01:19:26,000
used was a pressure plate sensor these

1875
01:19:23,840 --> 01:19:27,400
are on the feet of the robot and it's

1876
01:19:26,000 --> 01:19:28,600
basically just two metal plates and when

1877
01:19:27,400 --> 01:19:31,320
you connect them it makes a closed

1878
01:19:28,600 --> 01:19:33,360
circuit it's pretty simple um this allow

1879
01:19:31,320 --> 01:19:34,679
this tells the robot if its Feats feet

1880
01:19:33,360 --> 01:19:36,840
are touching the ground which is like

1881
01:19:34,679 --> 01:19:41,960
extra input for

1882
01:19:36,840 --> 01:19:45,000
learning um the robot has three arduinos

1883
01:19:41,960 --> 01:19:47,239
and one razy piie the first Arduino is

1884
01:19:45,000 --> 01:19:50,400
connected to the mpu 6050 for

1885
01:19:47,239 --> 01:19:52,719
orientation data the second Arduino is

1886
01:19:50,400 --> 01:19:55,320
connected to seros and pressure plate

1887
01:19:52,719 --> 01:19:57,960
data and the third U know is connected

1888
01:19:55,320 --> 01:19:59,920
to my battery backup power supply um all

1889
01:19:57,960 --> 01:20:02,440
it really is is just you can connect two

1890
01:19:59,920 --> 01:20:05,280
batteries it uses a voltage of Auto to

1891
01:20:02,440 --> 01:20:06,800
find how what voltage a battery is at

1892
01:20:05,280 --> 01:20:10,400
and if it's too low it switches to the

1893
01:20:06,800 --> 01:20:13,280
next battery it's pretty simple um all

1894
01:20:10,400 --> 01:20:15,960
the adinos are except for the third uino

1895
01:20:13,280 --> 01:20:18,040
is connected to the Raspberry Pi uh

1896
01:20:15,960 --> 01:20:21,480
which provides the environment data to

1897
01:20:18,040 --> 01:20:25,000
the pi which runs the td3 algorithm and

1898
01:20:21,480 --> 01:20:27,679
I also built this rig out of PVC pipes

1899
01:20:25,000 --> 01:20:32,040
cuz one this robot is not indestructible

1900
01:20:27,679 --> 01:20:35,760
so yeah um and it also helps it

1901
01:20:32,040 --> 01:20:38,760
train uh this is the legs demo

1902
01:20:35,760 --> 01:20:41,520
the yeah I the safety minis wouldn't let

1903
01:20:38,760 --> 01:20:44,239
me allow it to wouldn't allow me to

1904
01:20:41,520 --> 01:20:47,480
power on the thing cuz it's again it's a

1905
01:20:44,239 --> 01:20:49,120
ATX power supply from the dump so they

1906
01:20:47,480 --> 01:20:52,719
so they wouldn't let me power it on but

1907
01:20:49,120 --> 01:20:56,440
there is a video of it working and yeah

1908
01:20:52,719 --> 01:20:56,440
that's all you get

1909
01:20:57,740 --> 01:21:06,480
[Applause]

1910
01:21:03,280 --> 01:21:08,400
all right can can they hear me all right

1911
01:21:06,480 --> 01:21:10,840
all right uh so my name is helus and

1912
01:21:08,400 --> 01:21:12,760
I've um building the confetti dispenser

1913
01:21:10,840 --> 01:21:14,679
part for the project it was supposed to

1914
01:21:12,760 --> 01:21:17,040
go on top of the legs but you know

1915
01:21:14,679 --> 01:21:19,800
didn't right so we're going to quickly

1916
01:21:17,040 --> 01:21:22,239
go through the build of my project so uh

1917
01:21:19,800 --> 01:21:24,960
first thing is uh there's a reloading

1918
01:21:22,239 --> 01:21:27,600
mechanism which uh would be this thing

1919
01:21:24,960 --> 01:21:30,280
if uh it didn't get damaged during

1920
01:21:27,600 --> 01:21:32,000
transport uh basically rotates the sh so

1921
01:21:30,280 --> 01:21:35,000
that new shells can be loaded in the

1922
01:21:32,000 --> 01:21:37,679
chamber and then yeah you're not

1923
01:21:35,000 --> 01:21:39,480
propelled hey and then next we have the

1924
01:21:37,679 --> 01:21:41,600
controller box which is this thing right

1925
01:21:39,480 --> 01:21:44,840
here right it's pretty cool uh there's a

1926
01:21:41,600 --> 01:21:47,360
two fans on it and uh one giant uh

1927
01:21:44,840 --> 01:21:50,159
switch to turn the power on and off look

1928
01:21:47,360 --> 01:21:52,239
on the inside oh and it just fell apart

1929
01:21:50,159 --> 01:21:55,080
should be all right hey uh I'm not sure

1930
01:21:52,239 --> 01:21:58,639
if you can see it very much looks like a

1931
01:21:55,080 --> 01:22:01,159
bomb but I guarantee you it is

1932
01:21:58,639 --> 01:22:03,199
not it was uh pretty hard getting that

1933
01:22:01,159 --> 01:22:06,199
through

1934
01:22:03,199 --> 01:22:06,199
security

1935
01:22:06,320 --> 01:22:11,199
yeah now now we have the actual

1936
01:22:08,880 --> 01:22:13,760
dispenser unit right so it features two

1937
01:22:11,199 --> 01:22:16,679
flyws and a linear actuator to push it

1938
01:22:13,760 --> 01:22:19,840
forward so it actually launches and just

1939
01:22:16,679 --> 01:22:22,080
doesn't sit there right and for GUI it

1940
01:22:19,840 --> 01:22:24,560
is programmed in Python uh which allows

1941
01:22:22,080 --> 01:22:26,960
for the controls such as the safety uh

1942
01:22:24,560 --> 01:22:30,280
Auto aiming and voice

1943
01:22:26,960 --> 01:22:33,679
control oh that was not what happen

1944
01:22:30,280 --> 01:22:35,440
okay yeah okay and for joystick aiming I

1945
01:22:33,679 --> 01:22:37,639
used the py game library to receive an

1946
01:22:35,440 --> 01:22:40,480
input from the Logitech flight stick

1947
01:22:37,639 --> 01:22:43,960
which is uh down there cuz there's not

1948
01:22:40,480 --> 01:22:48,920
enough room on the table uh it was like

1949
01:22:43,960 --> 01:22:51,120
$40 uh some shady website but yeah it it

1950
01:22:48,920 --> 01:22:53,639
works though and for the input Supply I

1951
01:22:51,120 --> 01:22:56,560
decided to use 12 volts at 2.8 amps for

1952
01:22:53,639 --> 01:22:59,920
the flywheel actuator step promotor and

1953
01:22:56,560 --> 01:23:01,600
the servo uh I was stuck at uh 12 FS

1954
01:22:59,920 --> 01:23:04,080
because my teacher said uh no more than

1955
01:23:01,600 --> 01:23:06,400
12 FS so yeah and I would uh fail the

1956
01:23:04,080 --> 01:23:09,639
assignment so

1957
01:23:06,400 --> 01:23:12,360
yeah all right for the infrared tracking

1958
01:23:09,639 --> 01:23:17,000
component uh it's pretty simple I bought

1959
01:23:12,360 --> 01:23:20,480
like a infrared light filter off eBay it

1960
01:23:17,000 --> 01:23:22,679
was like pretty cheap 40 bux uh the

1961
01:23:20,480 --> 01:23:25,840
range is not very good but uh gets the

1962
01:23:22,679 --> 01:23:29,360
job done like TW not 20 meters like two

1963
01:23:25,840 --> 01:23:31,840
meters that's a bit too much honestly uh

1964
01:23:29,360 --> 01:23:34,920
yeah it works by applying a mask over

1965
01:23:31,840 --> 01:23:37,120
the point which is detected uh which is

1966
01:23:34,920 --> 01:23:39,360
normally red and black because yeah

1967
01:23:37,120 --> 01:23:41,520
that's just the color of infrared light

1968
01:23:39,360 --> 01:23:43,760
and then we have a Contours The Contours

1969
01:23:41,520 --> 01:23:45,760
are like for detecting uh what the parts

1970
01:23:43,760 --> 01:23:48,280
actually are I'm not describing it very

1971
01:23:45,760 --> 01:23:50,719
well but uh yeah so it creates like a

1972
01:23:48,280 --> 01:23:53,080
box and then I can calculate the center

1973
01:23:50,719 --> 01:23:54,560
using a bunch of math so and then I

1974
01:23:53,080 --> 01:23:57,880
would normally St values between between

1975
01:23:54,560 --> 01:24:00,360
negative 1 and one so uh the turret can

1976
01:23:57,880 --> 01:24:01,679
aim in its respective Direction like the

1977
01:24:00,360 --> 01:24:04,560
asmf and the

1978
01:24:01,679 --> 01:24:08,199
elevation right for the GUI it uses the

1979
01:24:04,560 --> 01:24:11,400
libraries pame CV2 tter custom tter

1980
01:24:08,199 --> 01:24:13,560
numpy n numpy not numpy uh it features

1981
01:24:11,400 --> 01:24:15,440
an Aiming display in the top right uh

1982
01:24:13,560 --> 01:24:17,480
button controls on the left side and

1983
01:24:15,440 --> 01:24:19,880
button controls include Auto aiming

1984
01:24:17,480 --> 01:24:22,920
safety and voice control so you can

1985
01:24:19,880 --> 01:24:24,920
toggle them on and off and then I've

1986
01:24:22,920 --> 01:24:28,159
utilized a multi threading system for

1987
01:24:24,920 --> 01:24:30,280
the GUI camera control and voice control

1988
01:24:28,159 --> 01:24:32,520
because they all operate at different uh

1989
01:24:30,280 --> 01:24:34,840
timings like for the camera it needs to

1990
01:24:32,520 --> 01:24:37,000
be like decently fast but not too fast

1991
01:24:34,840 --> 01:24:41,239
the GUI who cares it can be like two

1992
01:24:37,000 --> 01:24:42,880
frames per second which yeah it works

1993
01:24:41,239 --> 01:24:45,840
and controller needs to be as fast as

1994
01:24:42,880 --> 01:24:47,560
possible so it actually works and then

1995
01:24:45,840 --> 01:24:50,760
voice control cuz it's just on a wild

1996
01:24:47,560 --> 01:24:54,800
Loop and it wouldn't work all right so I

1997
01:24:50,760 --> 01:24:57,239
have a actual demo here is that sound no

1998
01:24:54,800 --> 01:24:59,800
right uh this was uh before I went on

1999
01:24:57,239 --> 01:25:01,960
the flight and it still

2000
01:24:59,800 --> 01:25:05,280
worked well just me uh testing the

2001
01:25:01,960 --> 01:25:05,280
flight stick you know pretty

2002
01:25:06,520 --> 01:25:11,600
cool uh it gets jammed on a wise a bit

2003
01:25:09,560 --> 01:25:13,800
because I'm not very good at a wire

2004
01:25:11,600 --> 01:25:17,040
management Cable

2005
01:25:13,800 --> 01:25:18,760
Management right yeah you can uh in the

2006
01:25:17,040 --> 01:25:19,800
back you're stuck with wires which I

2007
01:25:18,760 --> 01:25:22,360
found out

2008
01:25:19,800 --> 01:25:23,320
later it goes up and down you know it's

2009
01:25:22,360 --> 01:25:25,760
very

2010
01:25:23,320 --> 01:25:28,080
basic me uh questioning why it doesn't

2011
01:25:25,760 --> 01:25:28,080
work

2012
01:25:28,400 --> 01:25:35,280
here hey any more now uh we'll get to

2013
01:25:32,320 --> 01:25:35,280
good part soon trust

2014
01:25:35,480 --> 01:25:41,440
me it's long I know it's pretty long but

2015
01:25:38,760 --> 01:25:43,520
like will'll get there eventually I I

2016
01:25:41,440 --> 01:25:46,480
would skip ahead but I don't know how uh

2017
01:25:43,520 --> 01:25:49,119
canva works all right so now it actually

2018
01:25:46,480 --> 01:25:54,880
shows you the guy

2019
01:25:49,119 --> 01:25:57,159
right for for like0 two seconds but like

2020
01:25:54,880 --> 01:25:59,760
I'll show you g

2021
01:25:57,159 --> 01:26:02,400
later right let's me uh showing off like

2022
01:25:59,760 --> 01:26:03,840
the chamber and the linear actuator that

2023
01:26:02,400 --> 01:26:06,080
would push a shell forward I haven't

2024
01:26:03,840 --> 01:26:09,320
loaded a shell yet cuz uh there was

2025
01:26:06,080 --> 01:26:11,280
still like an assignment video so

2026
01:26:09,320 --> 01:26:14,600
yeah all

2027
01:26:11,280 --> 01:26:17,360
right was me uh being being a little

2028
01:26:14,600 --> 01:26:18,800
stupid and uh didn't notice that it

2029
01:26:17,360 --> 01:26:22,639
wasn't

2030
01:26:18,800 --> 01:26:24,480
loaded and uh here you go nor normally

2031
01:26:22,639 --> 01:26:27,080
there would be like a confetti in there

2032
01:26:24,480 --> 01:26:29,920
but uh I didn't load it cuz uh I just

2033
01:26:27,080 --> 01:26:32,719
kind of forgot I'll be

2034
01:26:29,920 --> 01:26:35,800
honest

2035
01:26:32,719 --> 01:26:38,800
yeah know this part uh yeah this the

2036
01:26:35,800 --> 01:26:41,040
infrared actual aiming part right so you

2037
01:26:38,800 --> 01:26:44,040
see the Red Dot that's like the infrared

2038
01:26:41,040 --> 01:26:46,880
light emitted from the flashlight it

2039
01:26:44,040 --> 01:26:49,000
automatically uh readjusts uh stops all

2040
01:26:46,880 --> 01:26:51,520
inputs from the

2041
01:26:49,000 --> 01:26:52,880
joystick right it should go down and

2042
01:26:51,520 --> 01:26:55,600
show

2043
01:26:52,880 --> 01:26:58,960
you any moment

2044
01:26:55,600 --> 01:27:03,440
now hey there

2045
01:26:58,960 --> 01:27:03,440
go oh there we go look how cute it

2046
01:27:03,639 --> 01:27:09,040
is it's

2047
01:27:06,280 --> 01:27:11,400
nodding that's enough of my project and

2048
01:27:09,040 --> 01:27:15,000
now we're going to go for uh to Matt

2049
01:27:11,400 --> 01:27:17,800
right yeah so I did the voice controls

2050
01:27:15,000 --> 01:27:19,440
and the screen for the project overall

2051
01:27:17,800 --> 01:27:20,840
um so there were two options for The

2052
01:27:19,440 --> 01:27:22,679
Voice Control you could either have it

2053
01:27:20,840 --> 01:27:24,320
controlling the confetti launcher or you

2054
01:27:22,679 --> 01:27:25,760
could have it as a virtual friend while

2055
01:27:24,320 --> 01:27:27,199
controlling the confetti launcher it

2056
01:27:25,760 --> 01:27:29,560
would use speech recognition python

2057
01:27:27,199 --> 01:27:31,480
Library through the Google speech API

2058
01:27:29,560 --> 01:27:33,080
which then looked for your keywords

2059
01:27:31,480 --> 01:27:34,679
which are the directions up down left

2060
01:27:33,080 --> 01:27:36,960
and right it looked for dispense or it

2061
01:27:34,679 --> 01:27:38,480
looked for voice control on or off uh if

2062
01:27:36,960 --> 01:27:40,159
you didn't have a voice control on for

2063
01:27:38,480 --> 01:27:43,040
the confetti launcher it would go to a

2064
01:27:40,159 --> 01:27:45,480
virtual friend which uses a llama 3.2

2065
01:27:43,040 --> 01:27:47,719
model to engage in conversation with the

2066
01:27:45,480 --> 01:27:49,119
user um uses a python text to speech

2067
01:27:47,719 --> 01:27:51,000
library to actually talk back to you so

2068
01:27:49,119 --> 01:27:52,320
it's not just text and while it's in

2069
01:27:51,000 --> 01:27:54,800
this mode the tracking is done via

2070
01:27:52,320 --> 01:27:56,040
helios's infrared system

2071
01:27:54,800 --> 01:27:58,239
uh these are just to show the commands

2072
01:27:56,040 --> 01:28:00,360
you know up down left right dispense

2073
01:27:58,239 --> 01:28:02,199
voice controls uh it was good to have it

2074
01:28:00,360 --> 01:28:03,400
be able to turn itself off because then

2075
01:28:02,199 --> 01:28:07,239
you know you don't have to go and touch

2076
01:28:03,400 --> 01:28:09,159
the GUI um the screen was used for

2077
01:28:07,239 --> 01:28:10,480
emotions and to give feedback on the

2078
01:28:09,159 --> 01:28:12,400
action so the user knows what they're

2079
01:28:10,480 --> 01:28:16,520
doing um you don't really want to be too

2080
01:28:12,400 --> 01:28:18,440
close in case it launches at you so

2081
01:28:16,520 --> 01:28:22,000
um

2082
01:28:18,440 --> 01:28:24,560
yeah so all this was good at least it

2083
01:28:22,000 --> 01:28:26,280
was the plan until as my friend Helios

2084
01:28:24,560 --> 01:28:28,679
pointed out some things went wrong a

2085
01:28:26,280 --> 01:28:31,480
couple quotes we took from about maybe 3

2086
01:28:28,679 --> 01:28:35,480
hours ago was one why is this not H

2087
01:28:31,480 --> 01:28:39,520
having power two why did it

2088
01:28:35,480 --> 01:28:41,639
snap three why is it in inverted and

2089
01:28:39,520 --> 01:28:43,159
four three out of the four h Bridges

2090
01:28:41,639 --> 01:28:47,840
killed

2091
01:28:43,159 --> 01:28:49,880
itself a um okay let's be real um there

2092
01:28:47,840 --> 01:28:52,679
are a lot of setbacks for this project

2093
01:28:49,880 --> 01:28:54,800
um the first being when you leave your

2094
01:28:52,679 --> 01:28:57,480
legs in the car before you go to the

2095
01:28:54,800 --> 01:29:02,000
airport in the sun and the plastic 3D

2096
01:28:57,480 --> 01:29:05,480
filament ends up melting um here yeah it

2097
01:29:02,000 --> 01:29:07,600
was yep um second thing is when the

2098
01:29:05,480 --> 01:29:11,000
servo screws end up snapping from being

2099
01:29:07,600 --> 01:29:13,880
tight and too tight um kind of a big

2100
01:29:11,000 --> 01:29:16,440
setback that's why it's so much bigger

2101
01:29:13,880 --> 01:29:18,480
originally um third when airplane

2102
01:29:16,440 --> 01:29:22,119
handling may accidentally break part of

2103
01:29:18,480 --> 01:29:26,199
your 3D printing um

2104
01:29:22,119 --> 01:29:28,560
ouch uh four is when 3 your four h

2105
01:29:26,199 --> 01:29:31,520
Bridges blow up and we still don't know

2106
01:29:28,560 --> 01:29:34,560
why um and five when the feeding

2107
01:29:31,520 --> 01:29:34,560
mechanism ends up

2108
01:29:35,360 --> 01:29:39,920
snapping um to start with it was

2109
01:29:38,080 --> 01:29:42,000
originally taller and you can see here

2110
01:29:39,920 --> 01:29:43,360
in the video that we accidentally filmed

2111
01:29:42,000 --> 01:29:46,119
when we accidentally locked our group

2112
01:29:43,360 --> 01:29:46,119
member out of the

2113
01:29:46,280 --> 01:29:51,920
room um you can see using it uh self-

2114
01:29:49,040 --> 01:29:53,520
learning to then try and stand up um it

2115
01:29:51,920 --> 01:29:55,480
also used to be able to turn but as you

2116
01:29:53,520 --> 01:29:57,840
can probably see here it it no longer

2117
01:29:55,480 --> 01:30:00,840
does that

2118
01:29:57,840 --> 01:30:00,840
um

2119
01:30:02,520 --> 01:30:08,239
yeah and then it also used to show faces

2120
01:30:05,480 --> 01:30:10,440
until this uina was in quotes from my

2121
01:30:08,239 --> 01:30:12,199
teacher cannibalized to get the last bit

2122
01:30:10,440 --> 01:30:17,000
working like the very last part of our

2123
01:30:12,199 --> 01:30:17,880
project at least somewhat working um yes

2124
01:30:17,000 --> 01:30:19,440
oh

2125
01:30:17,880 --> 01:30:23,000
you

2126
01:30:19,440 --> 01:30:25,440
um I've been M okay I've been trying my

2127
01:30:23,000 --> 01:30:29,360
hardest to uh make this thing be alive

2128
01:30:25,440 --> 01:30:33,000
again since like uh 9 in the morning uh

2129
01:30:29,360 --> 01:30:35,480
it's very sketchy

2130
01:30:33,000 --> 01:30:39,119
uh

2131
01:30:35,480 --> 01:30:42,080
like let's see if this uh works today

2132
01:30:39,119 --> 01:30:45,040
probably oh that's an

2133
01:30:42,080 --> 01:30:48,239
arrow oh

2134
01:30:45,040 --> 01:30:48,239
yeah all

2135
01:30:49,560 --> 01:30:55,040
right how does this side of the room

2136
01:30:52,440 --> 01:30:57,480
feel I noticed some people ducking down

2137
01:30:55,040 --> 01:31:02,400
a bit no it should it should be all

2138
01:30:57,480 --> 01:31:07,520
right I I should be it should be all

2139
01:31:02,400 --> 01:31:11,080
right uh I don't know let's see if that

2140
01:31:07,520 --> 01:31:11,080
uh died on the way up to the

2141
01:31:11,440 --> 01:31:16,320
stage might might have

2142
01:31:14,679 --> 01:31:18,679
died

2143
01:31:16,320 --> 01:31:22,119
fortunately so you don't get

2144
01:31:18,679 --> 01:31:24,159
splashed all right uh due to unforeseen

2145
01:31:22,119 --> 01:31:26,360
circumstances there won't be a life demo

2146
01:31:24,159 --> 01:31:29,560
actually

2147
01:31:26,360 --> 01:31:31,960
a um and the biggest takeaway is like

2148
01:31:29,560 --> 01:31:35,239
what did we learn from this project

2149
01:31:31,960 --> 01:31:38,360
[Laughter]

2150
01:31:35,239 --> 01:31:40,719
um firstly is don't package things um

2151
01:31:38,360 --> 01:31:42,600
unsafely and even if it is safe don't

2152
01:31:40,719 --> 01:31:43,480
add an additional packet of wires in the

2153
01:31:42,600 --> 01:31:48,480
same

2154
01:31:43,480 --> 01:31:51,920
spot um secondly plastic in cars it um

2155
01:31:48,480 --> 01:31:54,000
it melts uh we learned that the hard way

2156
01:31:51,920 --> 01:31:57,159
uh three contrary to the popular

2157
01:31:54,000 --> 01:32:01,480
engineering belief duct tape can't solve

2158
01:31:57,159 --> 01:32:04,440
everything um especially not last minute

2159
01:32:01,480 --> 01:32:05,520
so uh fourth cooling components is very

2160
01:32:04,440 --> 01:32:07,040
important when you don't have enough

2161
01:32:05,520 --> 01:32:10,320
resistance or it just generally isn't

2162
01:32:07,040 --> 01:32:10,320
cool enough things tend to blow

2163
01:32:10,880 --> 01:32:16,159
up um fifth plastic is not that strong

2164
01:32:13,920 --> 01:32:17,239
so when you're messing around with it um

2165
01:32:16,159 --> 01:32:20,080
it can

2166
01:32:17,239 --> 01:32:22,239
snap uh and sixth have spare parts to

2167
01:32:20,080 --> 01:32:24,520
make last minute adjustments uh we

2168
01:32:22,239 --> 01:32:25,800
didn't have many so um this is the best

2169
01:32:24,520 --> 01:32:29,060
we

2170
01:32:25,800 --> 01:32:32,030
had thank you very

2171
01:32:29,060 --> 01:32:33,200
[Applause]

2172
01:32:32,030 --> 01:32:42,119
[Laughter]

2173
01:32:33,200 --> 01:32:45,199
[Applause]

2174
01:32:42,119 --> 01:32:47,520
much that was amazing okay well I'm a

2175
01:32:45,199 --> 01:32:49,320
little bit sad that we weren't shouting

2176
01:32:47,520 --> 01:32:50,440
confetti but I am glad that I still have

2177
01:32:49,320 --> 01:32:53,679
my

2178
01:32:50,440 --> 01:32:56,280
eyesight nothing caught fire we don't

2179
01:32:53,679 --> 01:32:58,239
have have to get the uh safety team to

2180
01:32:56,280 --> 01:33:01,000
make

2181
01:32:58,239 --> 01:33:03,239
a to to worry about our care and then

2182
01:33:01,000 --> 01:33:09,360
apologize to the venue that could have

2183
01:33:03,239 --> 01:33:13,560
been concerning um we have a few minutes

2184
01:33:09,360 --> 01:33:15,119
but not many for some questions so uh

2185
01:33:13,560 --> 01:33:16,560
students do you want to all kind of

2186
01:33:15,119 --> 01:33:19,320
stand up

2187
01:33:16,560 --> 01:33:22,000
and don't crowd around the robot that's

2188
01:33:19,320 --> 01:33:25,480
not a great idea let's let's let's kind

2189
01:33:22,000 --> 01:33:27,040
of space use some space out here as well

2190
01:33:25,480 --> 01:33:29,800
uh let's let's take the robot off the

2191
01:33:27,040 --> 01:33:32,719
stage I don't want to touch it I'm sorry

2192
01:33:29,800 --> 01:33:34,920
I just got a power supply from the

2193
01:33:32,719 --> 01:33:38,040
dump you didn't tell me this earlier

2194
01:33:34,920 --> 01:33:41,480
about the yeah no I I I feel like I

2195
01:33:38,040 --> 01:33:43,719
learned a lot of things during that talk

2196
01:33:41,480 --> 01:33:46,840
and yet I still have more questions at

2197
01:33:43,719 --> 01:33:50,400
the end of the talk than than during the

2198
01:33:46,840 --> 01:33:50,400
talk pass um

2199
01:33:50,960 --> 01:33:55,679
perfect oh yep components it's all right

2200
01:33:54,159 --> 01:33:59,510
all we're

2201
01:33:55,679 --> 01:34:02,600
good it y yep

2202
01:33:59,510 --> 01:34:02,600
[Music]

2203
01:34:09,000 --> 01:34:15,520
um

2204
01:34:11,080 --> 01:34:17,119
I I look I yeah I'm future me is really

2205
01:34:15,520 --> 01:34:20,480
concerned about the risk assessment that

2206
01:34:17,119 --> 01:34:22,280
I'm going to have to do next year but um

2207
01:34:20,480 --> 01:34:24,280
that's that's future

2208
01:34:22,280 --> 01:34:25,920
that's I just like like to say again

2209
01:34:24,280 --> 01:34:27,220
these are not my students I don't have

2210
01:34:25,920 --> 01:34:30,159
duty of

2211
01:34:27,220 --> 01:34:34,080
[Music]

2212
01:34:30,159 --> 01:34:39,480
care okay um

2213
01:34:34,080 --> 01:34:41,360
so um what questions do we all have also

2214
01:34:39,480 --> 01:34:43,880
while while you're thinking up questions

2215
01:34:41,360 --> 01:34:45,639
um talking talking about it being

2216
01:34:43,880 --> 01:34:47,320
difficult to get this through security

2217
01:34:45,639 --> 01:34:50,080
and convince people that it's not a bomb

2218
01:34:47,320 --> 01:34:54,040
it's basically a

2219
01:34:50,080 --> 01:34:56,320
bomb you're you're setting you're using

2220
01:34:54,040 --> 01:34:57,639
exploding conf that like that sounds

2221
01:34:56,320 --> 01:35:00,960
like yeah okay maybe that's a

2222
01:34:57,639 --> 01:35:05,360
technicality um uh where is the

2223
01:35:00,960 --> 01:35:09,159
microphone uh what what people have

2224
01:35:05,360 --> 01:35:11,679
questions okay hang on come

2225
01:35:09,159 --> 01:35:14,280
down that wasn't a question that was I

2226
01:35:11,679 --> 01:35:16,480
can be your that your a question now uh

2227
01:35:14,280 --> 01:35:17,760
but we've we've only got one microphone

2228
01:35:16,480 --> 01:35:20,360
so we're going to have to run it back

2229
01:35:17,760 --> 01:35:23,320
and forth to theud I'm happy to do the

2230
01:35:20,360 --> 01:35:26,080
running oh very far thank you Evan just

2231
01:35:23,320 --> 01:35:32,400
I need my fitness up today you're

2232
01:35:26,080 --> 01:35:36,800
welcome um for vev the bipedal robot

2233
01:35:32,400 --> 01:35:39,880
team do you uh after your experience

2234
01:35:36,800 --> 01:35:42,520
with robotics and the failures of

2235
01:35:39,880 --> 01:35:44,199
mechanics in the real world do you enjoy

2236
01:35:42,520 --> 01:35:48,360
it more or

2237
01:35:44,199 --> 01:35:52,000
less think think com I'm com com yep yep

2238
01:35:48,360 --> 01:35:55,560
this is like a really bad relay okay all

2239
01:35:52,000 --> 01:35:57,320
right um is it enjoyable yes I mean uh

2240
01:35:55,560 --> 01:36:00,119
is it enjoy enjoyable for my bank

2241
01:35:57,320 --> 01:36:00,119
account not really

2242
01:36:02,719 --> 01:36:08,360
but uh it was pretty fun I mean it just

2243
01:36:05,639 --> 01:36:10,880
took way too long like some librar just

2244
01:36:08,360 --> 01:36:13,360
didn't work for some reason I don't know

2245
01:36:10,880 --> 01:36:14,600
why yeah staying up till like 2: a.m. to

2246
01:36:13,360 --> 01:36:16,560
download a bunch of libraries last

2247
01:36:14,600 --> 01:36:18,000
minute was not fun um I think all the

2248
01:36:16,560 --> 01:36:19,840
fun engineering Parts was actually

2249
01:36:18,000 --> 01:36:22,239
pretty good though probably do it more

2250
01:36:19,840 --> 01:36:25,040
from now on if it's

2251
01:36:22,239 --> 01:36:29,960
safe I feel like that is a loadbearing

2252
01:36:25,040 --> 01:36:32,280
if um are there any other

2253
01:36:29,960 --> 01:36:34,960
questions yes why don't you call and I

2254
01:36:32,280 --> 01:36:34,960
will repeat your

2255
01:36:35,119 --> 01:36:38,600
question question for

2256
01:36:41,159 --> 01:36:46,239
Annabelle question for Annabelle did you

2257
01:36:43,239 --> 01:36:48,960
think of using I need a better memory

2258
01:36:46,239 --> 01:36:52,760
the my God it's been a long day the XML

2259
01:36:48,960 --> 01:36:54,400
Library uh rather than Rex um I have not

2260
01:36:52,760 --> 01:36:57,320
heard of the XML library before but I

2261
01:36:54,400 --> 01:37:00,320
will look into that thank you for the

2262
01:36:57,320 --> 01:37:00,320
recommendation

2263
01:37:04,000 --> 01:37:09,440
uh a question for the

2264
01:37:06,760 --> 01:37:11,600
train let me let me repeat the question

2265
01:37:09,440 --> 01:37:15,600
um how long did you have to stand there

2266
01:37:11,600 --> 01:37:18,280
for the cargo train I filmed was only uh

2267
01:37:15,600 --> 01:37:21,880
2 minutes or so but but but the video

2268
01:37:18,280 --> 01:37:23,679
file is like 800 megabytes so it's it's

2269
01:37:21,880 --> 01:37:24,960
not small

2270
01:37:23,679 --> 01:37:32,360
that is a big

2271
01:37:24,960 --> 01:37:32,360
train um yes yes Nick um

2272
01:37:38,719 --> 01:37:42,679
the that that was a question of for the

2273
01:37:41,239 --> 01:37:45,599
wi-fi hacking was it you who had the

2274
01:37:42,679 --> 01:37:48,520
free candy network available earlier

2275
01:37:45,599 --> 01:37:50,840
today no that was not me but I want to

2276
01:37:48,520 --> 01:37:50,840
know who

2277
01:37:52,239 --> 01:37:57,920
did um

2278
01:37:54,440 --> 01:38:01,239
um if there are more

2279
01:37:57,920 --> 01:38:04,480
questions you can maybe oh there's one

2280
01:38:01,239 --> 01:38:04,480
okay yep shout it

2281
01:38:06,480 --> 01:38:12,360
out it's a compliment ah a compliment

2282
01:38:10,760 --> 01:38:14,520
for Annabelle hang

2283
01:38:12,360 --> 01:38:16,400
on hang on hang on this sounds like it

2284
01:38:14,520 --> 01:38:18,800
needs the microphone where's the

2285
01:38:16,400 --> 01:38:22,199
microphone it's a I know it I know it's

2286
01:38:18,800 --> 01:38:24,599
a comment I don't you could you could

2287
01:38:22,199 --> 01:38:26,119
phrase your question as a question by

2288
01:38:24,599 --> 01:38:29,480
asking everyone by complimenting and

2289
01:38:26,119 --> 01:38:31,920
then saying do you agree room and then

2290
01:38:29,480 --> 01:38:34,400
we could all

2291
01:38:31,920 --> 01:38:37,080
clap um basically for Annabel you know

2292
01:38:34,400 --> 01:38:38,520
what she has done is amazing right so I

2293
01:38:37,080 --> 01:38:40,840
work with the team you know to develop

2294
01:38:38,520 --> 01:38:42,560
applications and we look into the user

2295
01:38:40,840 --> 01:38:45,920
interface and you know um user

2296
01:38:42,560 --> 01:38:47,719
experience and then that's how in an

2297
01:38:45,920 --> 01:38:50,080
application in commercial scale or in

2298
01:38:47,719 --> 01:38:52,320
Enterprise scale what we can do is

2299
01:38:50,080 --> 01:38:54,679
utilize you know similar logic to

2300
01:38:52,320 --> 01:38:57,599
actually give you know look and feel of

2301
01:38:54,679 --> 01:38:59,520
what they like and look you know so ask

2302
01:38:57,599 --> 01:39:01,840
them to you know upload image you like

2303
01:38:59,520 --> 01:39:03,599
and then render the colors you know in

2304
01:39:01,840 --> 01:39:05,239
the in the scheme there you know um

2305
01:39:03,599 --> 01:39:07,560
based on your logic and that's really

2306
01:39:05,239 --> 01:39:09,639
great and so I can see a lot of you know

2307
01:39:07,560 --> 01:39:11,159
Brilliant Minds over here so yeah thank

2308
01:39:09,639 --> 01:39:17,520
you you know thank you very much

2309
01:39:11,159 --> 01:39:17,520
everybody so room do you agree yeah

2310
01:39:17,760 --> 01:39:25,360
yeah hey room do you agree that all of

2311
01:39:21,199 --> 01:39:25,360
our students are amazing

2312
01:39:28,400 --> 01:39:35,520
I do

2313
01:39:30,760 --> 01:39:38,159
um okay I will grab some slides out can

2314
01:39:35,520 --> 01:39:40,080
I get my co-presenters to give all of

2315
01:39:38,159 --> 01:39:42,360
the students a

2316
01:39:40,080 --> 01:39:46,840
mug and now I'm going to make the

2317
01:39:42,360 --> 01:39:46,840
students act out their own favorite

2318
01:39:49,920 --> 01:39:53,920
emoji um

2319
01:39:56,600 --> 01:40:01,120
[Laughter]

2320
01:40:01,360 --> 01:40:05,880
can we get one more massive Round of

2321
01:40:02,719 --> 01:40:09,360
Applause for our fabulous

2322
01:40:05,880 --> 01:40:12,560
students put it down go on shoot go done

2323
01:40:09,360 --> 01:40:12,560
you're done