1 of 21

1

Cat or Dog?

Hands-on with CNNs

Matthew Kenely

2 of 21

2

What is a cat?

Pointy ears

Whiskers

Fur

.

.

.

*Probably* a cat…

3 of 21

3

FEATURES

4 of 21

4

How can a computer see cats?

If a picture contains lots of features typical of a cat, then it is probably a picture of a cat.

Assumption

5 of 21

5

AI

Machine Learning

Deep Learning

CNNs

Linear Regression

Decision Trees

SVMs

...

Artificial Neural Networks

Recurrent Neural Networks

...

Computer Vision

SURF, SIFT, Harris

Canny, Sobel

Gaussian, median

Dilation, erosion

6 of 21

6

Convolutional Neural Networks

7 of 21

7

Convolutional Neural Networks

8 of 21

8

Cat? meter

9 of 21

9

Cat? meter

10 of 21

10

Cat? meter

11 of 21

11

Cat? meter

12 of 21

12

Cat? meter

13 of 21

13

Cat? meter

14 of 21

14

Cat? meter

15 of 21

15

16 of 21

16

{

{

226 152 207

243 226 210

238 177 152

226 157 182

117 191 191

209 238 123

212 209 151

176 111 247

251 107 228

153 207 140

178 234 136

209 195 125

184 239 197

232 114 207

241 114 219

109 235 115

167 220 102

255 242 204

208 187 115

114 137 250

180 215 252

111 138 214

150 110 172

175 138 162

136 124 230

129 223 241

162 174 219

17 of 21

17

Convolution

1

1

1

0

0

0

1

1

1

18 of 21

18

Convolution

19 of 21

19

Convolution

20 of 21

20

mkenely.com/cat-or-dog

21 of 21

21

Cat or Dog?

Hands-on with CNNs

Matthew Kenely • mkenely.com