BBM413

Fundamentals of
Image Processing

Convolutional Neural Networks

Erkut Erdem
Hacettepe University
Computer Vision Lab (HUCVL)
 Convolution Layer
32x32x3 image

32 height
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32 width
3 depth

!2
 Convolution Layer
32x32x3 image

5x5x3 filter
32

Convolve the filter with the

image
i.e. “slide over the image
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

spatially, computing dot

32
products”
3

!3
 Convolution Layer
Filters always extend the full
depth of the input volume
32x32x3 image

5x5x3 filter
32

Convolve the filter with the

image
i.e. “slide over the image
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

spatially, computing dot

32
products”
3

!4
 Convolution Layer
32x32x3 image
5x5x3 filter
32

1 number:
the result of taking a dot product between the
filter and a small 5x5x3 chunk of the image
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32 (i.e. 5*5*3 = 75-dimensional dot product + bias)

3

!5
 Convolution Layer
activation
32x32x3 image map
5x5x3 filter
32

28

convolve (slide) over all

spatial locations
28
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32
3 1

!6
 Convolution Layer
consider a second, green filter
32x32x3 image activation maps

5x5x3 filter
32

28

convolve (slide) over all

spatial locations
28
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32
3 1

!7
 For example, if we had 6 5x5 filters, we’ll get 6
separate activation maps:
activation maps

32

28

Convolution Layer

32 28
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

3 6

We stack these up to get a “new image” of size 28x28x6!

!8
 Preview: ConvNet is a sequence of
Convolutional Layers, interspersed with
activation functions

32 28

CONV,
ReLU
e.g. 6
5x5x3
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32 28
filters
3 6

!9
 Preview: ConvNet is a sequence of
Convolutional Layers, interspersed with
activation functions

32 28 24

….
CONV, CONV, CONV,
ReLU ReLU ReLU
e.g. 6 e.g. 10
5x5x3 5x5x6
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32 28 24
filters filters
3 6 10

!10
 !11
[From recent Yann
LeCun slides]
Preview
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 !12
[From recent Yann
LeCun slides]
Preview
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 one filter =>
one activation map
example 5x5
filters
(32 total)
We call the layer convolutional
because it is related to
convolution of two signals:
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

elementwise multiplication
and sum of a filter and the
signal (image)

!13
 !14
!14
Preview
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 A closer look at spatial
dimensions:
activation
32x32x3 image map
5x5x3 filter
32

28

convolve (slide) over all

spatial locations
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32 28
3 1

!15
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter

7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!16
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter

7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!17
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter

7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!18
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter

7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!19
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter

=> 5x5 output

7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!20
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter
applied with stride 2

7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!21
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter
applied with stride 2
7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!22
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter
applied with stride 2
=> 3x3 output!
7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!23
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter
applied with stride 3?
7
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!24
 A closer look at spatial
dimensions:
7
7x7 input (spatially)
assume 3x3 filter
applied with stride 3?
7
doesn’t fit!
cannot apply 3x3 filter on
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

7x7 input with stride 3.

!25
 N
Output size:
(N - F) / stride + 1
F
e.g. N = 7, F = 3:
F N stride 1 => (7 - 3)/1 + 1 = 5
stride 2 => (7 - 3)/2 + 1 = 3
stride 3 => (7 - 3)/3 + 1 = 2.33 :
\
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!26
 In practice: Common to zero pad
the border
0 0 0 0 0 0 e.g. input 7x7
3x3 filter, applied with stride 1
0
pad with 1 pixel border => what is the
0 output?

0
0
(recall:)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

(N - F) / stride + 1

!27
 In practice: Common to zero pad
the border
0 0 0 0 0 0 e.g. input 7x7
3x3 filter, applied with stride 1
0
pad with 1 pixel border => what is the
0 output?

0 7x7 output!
0
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!28
 In practice: Common to zero pad
the border
0 0 0 0 0 0 e.g. input 7x7
3x3 filter, applied with stride 1
0
pad with 1 pixel border => what is the
0 output?

0 7x7 output!
0 in general, common to see CONV layers
with stride 1, filters of size FxF, and zero-
padding with (F-1)/2. (will preserve size
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

spatially)
e.g. F = 3 => zero pad with 1
F = 5 => zero pad with 2
F = 7 => zero pad with 3

!29
 Remember back to…
E.g. 32x32 input convolved repeatedly with 5x5 filters
shrinks volumes spatially!
(32 -> 28 -> 24 ...). Shrinking too fast is not good, doesn’t
work well.

32 28 24

….
CONV, CONV, CONV,
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

ReLU ReLU ReLU

e.g. 6 e.g. 10
5x5x3 5x5x6
32 28 24
filters filters
3 6 10

!30
 Recap: Convolution Layer

(No padding, no strides)

Convolving a 3 × 3 kernel over a 4 × 4 input using unit strides
(i.e., i = 4, k = 3, s = 1 and p = 0).

Image credit: Vincent Dumoulin and Francesco Visin !31

Computing the output values of a 2D discrete convolution
i1 = i2 = 5, k1 = k2 = 3, s1 = s2 = 2, and p1 = p2 = 1

Image credit: Vincent Dumoulin and Francesco Visin !32

 Examples
time:
Input volume: 32x32x3
10 5x5 filters with stride 1, pad 2

Output volume size: ?

slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!33
 Examples
time:
Input volume: 32x32x3
10 5x5 filters with stride 1, pad 2

Output volume size:

(32+2*2-5)/1+1 = 32 spatially, so
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32x32x10

!34
 Examples
time:
Input volume: 32x32x3
10 5x5 filters with stride 1, pad 2

Number of parameters in this layer?

slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!35
 Examples
time:
Input volume: 32x32x3
10 5x5 filters with stride 1, pad 2

Number of parameters in this layer?

each filter has 5*5*3 + 1 = 76 params (+1 for bias)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

=> 76*10 = 760

!36
 !37
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 Common settings:

K = (powers of 2, e.g. 32, 64, 128,

512)
- F = 3, S = 1, P = 1
- F = 5, S = 1, P = 2
- F = 5, S = 2, P = ? (whatever fits)
- F = 1, S = 1, P = 0
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!38
 (btw, 1x1 convolution layers make perfect
sense)

1x1 CONV
56 with 32 filters
56
(each filter has size
1x1x64, and performs a
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

64-dimensional dot
56 product) 56
64 32

!39
 The brain/neuron view of CONV
Layer

32x32x3 image
5x5x3 filter
32
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

1 number:
32 the result of taking a dot product between
3 the filter and this part of the image
(i.e. 5*5*3 = 75-dimensional dot product)

!40
 The brain/neuron view of CONV
Layer

32x32x3 image
5x5x3 filter
32

It’s just a neuron with local

connectivity...
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

1 number:
32 the result of taking a dot product between
3 the filter and this part of the image
(i.e. 5*5*3 = 75-dimensional dot product)

!41
 The brain/neuron view of CONV
Layer

32

28 An activation map is a 28x28 sheet of neuron

outputs:
1. Each is connected to a small region in the input
2. All of them share parameters
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

32
28 “5x5 filter” -> “5x5 receptive field for each neuron”
3

!42
 The brain/neuron view of CONV
Layer

32

E.g. with 5 filters,

28 CONV layer consists of
neurons arranged in a 3D grid
(28x28x5)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

There will be 5 different

32 28 neurons all looking at the
3 5 same region in the input
volume

!43
 !44
Activation Functions
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 Activation Functions

Sigmoid

tanh tanh(x)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

ReLU max(0,x)
!45
 Activation Functions

- Squashes numbers to range [0,1]

- Historically popular since they
have nice interpretation as a
saturating “firing rate” of a neuron

3 problems:

1. Saturated neurons “kill” the

slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Sigmoid gradients
2. Sigmoid outputs are not zero-
centered
3. exp() is a bit compute expensive

!46
 Activation Functions

- Squashes numbers to range [-1,1]

- zero centered (nice)
- still kills gradients when saturated :(

tanh(x)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[LeCun et al., 1991]

!47
 Activation Functions - Computes f(x) = max(0,x)
- Does not saturate (in
+region)
- Very computationally
efficient
- Converges much faster than
sigmoid/tanh in practice
(e.g. 6x)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

ReLU
(Rectified Linear Unit)

[Krizhevsky et al., 2012]

!48
 !49
!49
two more layers to go: POOL/FC
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 Pooling layer
- makes the representations smaller and more manageable
- operates over each activation map independently:
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!50
 Max Pooling

Single depth
slice
1 1 2 4
x max pool with 2x2
5 6 7 8 filters and stride 2 6 8

3 2 1 0 3 4

1 2 3 4
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!51
 !52
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 !53
Common settings:

F = 2, S = 2
F = 3, S = 2
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 Fully Connected Layer (FC layer)
- Contains neurons that connect to the entire input volume, as in ordinary
Neural Networks
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!54
!54
 [ConvNetJS demo: training on CIFAR-10]

http://cs.stanford.edu/people/karpathy/convnetjs/demo/cifar10.html
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!55
Case studies

!56
 Case Study: LeNet-5 [LeCun et al., 1998]
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Conv filters were 5x5, applied at stride 1

Subsampling (Pooling) layers were 2x2 applied at stride
2
i.e. architecture is [CONV-POOL-CONV-POOL-CONV-
FC]
!57
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

First layer (CONV1): 96 11x11 filters applied at stride 4

=>
Q: what is the output volume size? Hint: (227-11)/4+1 = 55
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!58
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

First layer (CONV1): 96 11x11 filters applied at stride 4

=>
Output volume [55x55x96]
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Q: What is the total number of parameters in this layer?

!59
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

First layer (CONV1): 96 11x11 filters applied at stride 4

=>
Output volume [55x55x96]
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Parameters: (11*11*3)*96 = 35K

!60
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

After CONV1: 55x55x96

Second layer (POOL1): 3x3 filters applied at stride 2

slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Q: what is the output volume size? Hint: (55-3)/2+1 = 27

!61
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

After CONV1: 55x55x96

Second layer (POOL1): 3x3 filters applied at stride 2

Output volume: 27x27x96
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Q: what is the number of parameters in this layer?

!62
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

After CONV1: 55x55x96

Second layer (POOL1): 3x3 filters applied at stride 2

Output volume: 27x27x96
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Parameters: 0!

!63
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Input: 227x227x3 images

After CONV1: 55x55x96
After POOL1: 27x27x96
...
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!64
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Full (simplified) AlexNet architecture:

[227x227x3] INPUT
[55x55x96] CONV1: 96 11x11 filters at stride 4, pad 0
[27x27x96] MAX POOL1: 3x3 filters at stride 2
[27x27x96] NORM1: Normalization layer
[27x27x256] CONV2: 256 5x5 filters at stride 1, pad 2
[13x13x256] MAX POOL2: 3x3 filters at stride 2
[13x13x256] NORM2: Normalization layer
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[13x13x384] CONV3: 384 3x3 filters at stride 1, pad 1

[13x13x384] CONV4: 384 3x3 filters at stride 1, pad 1
[13x13x256] CONV5: 256 3x3 filters at stride 1, pad 1
[6x6x256] MAX POOL3: 3x3 filters at stride 2
[4096] FC6: 4096 neurons
[4096] FC7: 4096 neurons
[1000] FC8: 1000 neurons (class scores)
!65
 Case Study: AlexNet
[Krizhevsky et al. 2012]

Full (simplified) AlexNet architecture:

[227x227x3] INPUT
[55x55x96] CONV1: 96 11x11 filters at stride 4, pad 0 Details/Retrospectives:
[27x27x96] MAX POOL1: 3x3 filters at stride 2 - first use of ReLU
[27x27x96] NORM1: Normalization layer - used Norm layers (not common
[27x27x256] CONV2: 256 5x5 filters at stride 1, pad 2 anymore)
[13x13x256] MAX POOL2: 3x3 filters at stride 2 - heavy data augmentation
[13x13x256] NORM2: Normalization layer
- dropout 0.5
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[13x13x384] CONV3: 384 3x3 filters at stride 1, pad 1

- batch size 128
[13x13x384] CONV4: 384 3x3 filters at stride 1, pad 1
[13x13x256] CONV5: 256 3x3 filters at stride 1, pad 1 - SGD Momentum 0.9
[6x6x256] MAX POOL3: 3x3 filters at stride 2 - Learning rate 1e-2, reduced by 10
[4096] FC6: 4096 neurons manually when val accuracy plateaus
[4096] FC7: 4096 neurons - L2 weight decay 5e-4
[1000] FC8: 1000 neurons (class scores) - 7 CNN ensemble: 18.2% -> 15.4%
!66
 Case Study: VGGNet
[Simonyan and Zisserman, 2014]

Only 3x3 CONV stride 1, pad 1

and 2x2 MAX POOL stride 2

best model
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

11.2% top 5 error in ILSVRC

2013
->
7.3% top 5 error

!67
 INPUT: [224x224x3] memory: 224*224*3=150K params: 0
(not counting biases)
CONV3-64: [224x224x64] memory: 224*224*64=3.2M params: (3*3*3)*64 = 1,728
CONV3-64: [224x224x64] memory: 224*224*64=3.2M params: (3*3*64)*64 = 36,864
POOL2: [112x112x64] memory: 112*112*64=800K params: 0
CONV3-128: [112x112x128] memory: 112*112*128=1.6M params: (3*3*64)*128 = 73,728
CONV3-128: [112x112x128] memory: 112*112*128=1.6M params: (3*3*128)*128 = 147,456
POOL2: [56x56x128] memory: 56*56*128=400K params: 0
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*128)*256 = 294,912
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*256)*256 = 589,824
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*256)*256 = 589,824
POOL2: [28x28x256] memory: 28*28*256=200K params: 0
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*256)*512 = 1,179,648
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*512)*512 = 2,359,296
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*512)*512 = 2,359,296
POOL2: [14x14x512] memory: 14*14*512=100K params: 0
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

POOL2: [7x7x512] memory: 7*7*512=25K params: 0

FC: [1x1x4096] memory: 4096 params: 7*7*512*4096 = 102,760,448
FC: [1x1x4096] memory: 4096 params: 4096*4096 = 16,777,216
FC: [1x1x1000] memory: 1000 params: 4096*1000 = 4,096,000

!68
 (not counting biases)
INPUT: [224x224x3] memory: 224*224*3=150K params: 0
CONV3-64: [224x224x64] memory: 224*224*64=3.2M params: (3*3*3)*64 = 1,728
CONV3-64: [224x224x64] memory: 224*224*64=3.2M params: (3*3*64)*64 = 36,864
POOL2: [112x112x64] memory: 112*112*64=800K params: 0
CONV3-128: [112x112x128] memory: 112*112*128=1.6M params: (3*3*64)*128 = 73,728
CONV3-128: [112x112x128] memory: 112*112*128=1.6M params: (3*3*128)*128 = 147,456
POOL2: [56x56x128] memory: 56*56*128=400K params: 0
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*128)*256 = 294,912
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*256)*256 = 589,824
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*256)*256 = 589,824
POOL2: [28x28x256] memory: 28*28*256=200K params: 0
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*256)*512 = 1,179,648
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*512)*512 = 2,359,296
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*512)*512 = 2,359,296
POOL2: [14x14x512] memory: 14*14*512=100K params: 0
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

POOL2: [7x7x512] memory: 7*7*512=25K params: 0

FC: [1x1x4096] memory: 4096 params: 7*7*512*4096 = 102,760,448
FC: [1x1x4096] memory: 4096 params: 4096*4096 = 16,777,216
FC: [1x1x1000] memory: 1000 params: 4096*1000 = 4,096,000

TOTAL memory: 24M * 4 bytes ~= 93MB / image (only forward! ~*2 for bwd)
TOTAL params: 138M parameters

!69
 INPUT: [224x224x3] memory: 224*224*3=150K params: 0
(not counting biases) Note:
CONV3-64: [224x224x64] memory: 224*224*64=3.2M params: (3*3*3)*64 = 1,728
CONV3-64: [224x224x64] memory: 224*224*64=3.2M params: (3*3*64)*64 = 36,864
POOL2: [112x112x64] memory: 112*112*64=800K params: 0 Most memory is in
CONV3-128: [112x112x128] memory: 112*112*128=1.6M params: (3*3*64)*128 = 73,728 early CONV
CONV3-128: [112x112x128] memory: 112*112*128=1.6M params: (3*3*128)*128 = 147,456
POOL2: [56x56x128] memory: 56*56*128=400K params: 0
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*128)*256 = 294,912
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*256)*256 = 589,824
CONV3-256: [56x56x256] memory: 56*56*256=800K params: (3*3*256)*256 = 589,824
POOL2: [28x28x256] memory: 28*28*256=200K params: 0
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*256)*512 = 1,179,648
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*512)*512 = 2,359,296
CONV3-512: [28x28x512] memory: 28*28*512=400K params: (3*3*512)*512 = 2,359,296
POOL2: [14x14x512] memory: 14*14*512=100K params: 0
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
CONV3-512: [14x14x512] memory: 14*14*512=100K params: (3*3*512)*512 = 2,359,296
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

POOL2: [7x7x512] memory: 7*7*512=25K params: 0 Most params are

FC: [1x1x4096] memory: 4096 params: 7*7*512*4096 = 102,760,448
FC: [1x1x4096] memory: 4096 params: 4096*4096 = 16,777,216 in late FC
FC: [1x1x1000] memory: 1000 params: 4096*1000 = 4,096,000

TOTAL memory: 24M * 4 bytes ~= 93MB / image (only forward! ~*2 for bwd)
TOTAL params: 138M parameters

!70
 Case Study: GoogLeNet [Szegedy et al.,
2014]

Inception module
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

ILSVRC 2014 winner (6.7% top 5

error)

!71
 Case Study: ResNetILSVRC
[He et al., 2015]
2015 winner
(3.6% top 5 error)
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Slide from Kaiming He’s recent presentation https://www.youtube.com/

watch?v=1PGLj-uKT1w !72
 Case Study: ResNetILSVRC
[He et al., 2015]
2015 winner
(3.6% top 5 error)

2-3 weeks of
training on 8 GPU
machine

at runtime: faster
than a VGGNet!
(even though it has
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

8x more layers)

(slide from Kaiming He’s recent presentation)

!73
 Case Study: ResNet [He et al., 2015]

224x224x3

spatial dimension
only 56x56!
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!74
 Input
Input
ImageInput
Image 96 256 354
Image filters filters filters
7x7x3 Convolution 5x5x96 Convolution 3x3x256 Convolution
RGB Input Image 3x3 Max Pooling 3x3 Max Pooling 13 x 13 x 354
224 x 224 x 3 Down Sample 4x Down Sample 4x
55 x 55 x 96 13 x 13 x 256

256 354
filters filters
3x3x354 Convolution 3x3x354 Convolution
Logistic Standard Standard 3x3 Max Pooling
Regression 4096 Units 4096 Units 13 x 13 x 354
Down Sample 2x
≈1000 Classes 6 x 6 x 256
slide by Yisong Yue

http://www.image-net.org/ http://cs.nyu.edu/~fergus/papers/zeilerECCV2014.pdf
http://cs.nyu.edu/~fergus/presentations/nips2013_final.pdf
 Visualizing CNN (Layer 1)
slide by Yisong Yue

http://cs.nyu.edu/~fergus/papers/zeilerECCV2014.pdf
http://cs.nyu.edu/~fergus/presentations/nips2013_final.pdf
76
 Visualizing CNN (Layer 2)

Part that Triggered Filter Top Image Patches

slide by Yisong Yue

http://cs.nyu.edu/~fergus/papers/zeilerECCV2014.pdf
http://cs.nyu.edu/~fergus/presentations/nips2013_final.pdf
77
 Visualizing CNN (Layer 3)

Part that Triggered Filter Top Image Patches

slide by Yisong Yue

http://cs.nyu.edu/~fergus/papers/zeilerECCV2014.pdf
http://cs.nyu.edu/~fergus/presentations/nips2013_final.pdf
78
 Visualizing CNN (Layer 4)

Part that Triggered Filter Top Image Patches

slide by Yisong Yue

http://cs.nyu.edu/~fergus/papers/zeilerECCV2014.pdf
http://cs.nyu.edu/~fergus/presentations/nips2013_final.pdf
79
 Visualizing CNN (Layer 5)

Part that Triggered Filter Top Image Patches

slide by Yisong Yue

http://cs.nyu.edu/~fergus/papers/zeilerECCV2014.pdf
http://cs.nyu.edu/~fergus/presentations/nips2013_final.pdf
80
Tips and Tricks

!81
• Shuffle the training samples

• Use Dropoout and Batch

Normalization for regularization

!82
 “Given a rectangular image, we first rescaled

Input representation the image such that the shorter side was of
length 256, and then cropped out the central
256×256 patch from the resulting image”

• Centered (0-mean) RGB values.

slide by Alex Krizhevsky

!83
 Data Augmentation
• The neural net has 60M
real-valued parameters and
650,000 neurons
• It overfits a lot. Therefore,
they train on 224x224
patches extracted randomly
from 256x256 images, and
also their horizontal
reflections.
“This increases the size of our training
slide by Alex Krizhevsky

set by a factor of 2048, though the

resulting training examples are, of
course, highly inter- dependent.” [Krizhevsky et al. 2012] !84
 Data Augmentation
• Alter the intensities of the
RGB channels in training
images.
“Specifically, we perform PCA on the set
of RGB pixel values throughout the
ImageNet training set. To each training
image, we add multiples of the found
principal components, with magnitudes
proportional to the corres. ponding
eigenvalues times a random variable
drawn from a Gaussian with mean zero
and standard deviation 0.1…This scheme
approximately captures an important
property of natural images, namely, that
slide by Alex Krizhevsky

object identity is invariant to changes in

the intensity and color of the illumination.
This scheme reduces the top-1 error rate [Krizhevsky et al. 2012]
by over 1%.” !85
 !86
Data Augmentation
Horizontal flips
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson
 Data Augmentation
Get creative!

Random mix/combinations of :
- translation
- rotation
- stretching
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

- shearing,
- lens distortions, … (go crazy)

!87
 Transfer Learning with ConvNets

1. Train on
Imagenet
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

!88
 Transfer Learning with ConvNets
2. Small dataset:
1. Train on feature extractor
Imagenet

Freeze
these
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Train
this

!89
 Transfer Learning with ConvNets
2. Small dataset: 3. Medium dataset:
1. Train on feature extractor finetuning
Imagenet
more data = retrain more of
the network (or all of it)
Freeze these
Freeze
these
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Train Train this

this

!90
 Transfer Learning with ConvNets
2. Small dataset: 3. Medium dataset:
1. Train on feature extractor finetuning
Imagenet
more data = retrain more of
the network (or all of it)
Freeze these
Freeze tip: use only ~1/10th of
these the original learning rate
in finetuning top layer,
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

and ~1/100th on
intermediate layers

Train Train this

this

!91
 Today ConvNets are everywhere
Classification Retrieval
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Krizhevsky 2012]

!92
 Today ConvNets are everywhere
Detection Segmentation
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Faster R-CNN: Ren, He, Girshick, Sun 2015] [Farabet et al., 2012]

!93
 Today ConvNets are everywhere

NVIDIA Tegra X1
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

self-driving cars

!94
 Today ConvNets are everywhere
[Taigman et al.
2014]
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Simonyan et al. 2014] [Goodfellow 2014]

!95
 Today ConvNets are everywhere

[Toshev, Szegedy 2014]

slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Mnih 2013]

!96
 Today ConvNets are everywhere
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Ciresan et al. 2013] [Sermanet et al. 2011]

[Ciresan et al.]

!97
 Today ConvNets are everywhere

[Denil et al. 2014

slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Turaga et al., 2010]

!98
 Today ConvNets are everywhere
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

Whale recognition, Kaggle Challenge Mnih and Hinton, 2010

!99
 Today ConvNets are everywhere
Image
Captioning
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

[Vinyals et al., 2015]

!100
 Today ConvNets are everywhere
slide by Fei-Fei Li, Andrej Karpathy & Justin Johnson

reddit.com/r/deepdream

!101