-
Notifications
You must be signed in to change notification settings - Fork 334
/
cnn.py
106 lines (82 loc) · 4.37 KB
/
cnn.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
import numpy as np
import scipy.signal
def sigmoid(x):
return 1 / (1 + np.exp(-x))
def cnn_convolve(patch_dim, num_features, images, W, b, zca_white, patch_mean):
"""
Returns the convolution of the features given by W and b with
the given images
:param patch_dim: patch (feature) dimension
:param num_features: number of features
:param images: large images to convolve with, matrix in the form
images(r, c, channel, image number)
:param W: weights of the sparse autoencoder
:param b: bias of the sparse autoencoder
:param zca_white: zca whitening
:param patch_mean: mean of the images
:return:
"""
num_images = images.shape[3]
image_dim = images.shape[0]
image_channels = images.shape[2]
# Instructions:
# Convolve every feature with every large image here to produce the
# numFeatures x numImages x (imageDim - patchDim + 1) x (imageDim - patchDim + 1)
# matrix convolvedFeatures, such that
# convolvedFeatures(featureNum, imageNum, imageRow, imageCol) is the
# value of the convolved featureNum feature for the imageNum image over
# the region (imageRow, imageCol) to (imageRow + patchDim - 1, imageCol + patchDim - 1)
#
# Expected running times:
# Convolving with 100 images should take less than 3 minutes
# Convolving with 5000 images should take around an hour
# (So to save time when testing, you should convolve with less images, as
# described earlier)
convolved_features = np.zeros(shape=(num_features, num_images, image_dim - patch_dim + 1,
image_dim - patch_dim + 1),
dtype=np.float64)
WT = W.dot(zca_white)
bT = b - WT.dot(patch_mean)
for i in range(num_images):
for j in range(num_features):
# convolution of image with feature matrix for each channel
convolved_image = np.zeros(shape=(image_dim - patch_dim + 1, image_dim - patch_dim + 1),
dtype=np.float64)
for channel in range(image_channels):
# Obtain the feature (patchDim x patchDim) needed during the convolution
patch_size = patch_dim * patch_dim
feature = WT[j, patch_size * channel:patch_size * (channel + 1)].reshape(patch_dim, patch_dim)
# Flip the feature matrix because of the definition of convolution, as explained later
feature = np.flipud(np.fliplr(feature))
# Obtain the image
im = images[:, :, channel, i]
# Convolve "feature" with "im", adding the result to convolvedImage
# be sure to do a 'valid' convolution
convolved_image += scipy.signal.convolve2d(im, feature, mode='valid')
# Subtract the bias unit (correcting for the mean subtraction as well)
# Then, apply the sigmoid function to get the hidden activation
convolved_image = sigmoid(convolved_image + bT[j])
# The convolved feature is the sum of the convolved values for all channels
convolved_features[j, i, :, :] = convolved_image
return convolved_features
def cnn_pool(pool_dim, convolved_features):
"""
Pools the given convolved features
:param pool_dim: dimension of the pooling region
:param convolved_features: convolved features to pool (as given by cnn_convolve)
convolved_features(feature_num, image_num, image_row, image_col)
:return: pooled_features: matrix of pooled features in the form
pooledFeatures(featureNum, imageNum, poolRow, poolCol)
"""
num_images = convolved_features.shape[1]
num_features = convolved_features.shape[0]
convolved_dim = convolved_features.shape[2]
assert convolved_dim % pool_dim == 0, "Pooling dimension is not an exact multiple of convolved dimension"
pool_size = convolved_dim / pool_dim
pooled_features = np.zeros(shape=(num_features, num_images, pool_size, pool_size),
dtype=np.float64)
for i in range(pool_size):
for j in range(pool_size):
pool = convolved_features[:, :, i * pool_dim:(i + 1) * pool_dim, j * pool_dim:(j + 1) * pool_dim]
pooled_features[:, :, i, j] = np.mean(np.mean(pool, 2), 2)
return pooled_features