-
Notifications
You must be signed in to change notification settings - Fork 1
/
viz.py
127 lines (102 loc) · 4.04 KB
/
viz.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
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
import os
import csv
import time
import random
import numpy as np
from scipy.misc import imread, imsave
from keras.models import load_model
from keras import backend as K
model = load_model('./model.h5')
input_img = model.input
img_height, img_width = input_img.shape[1:3]
layer_dict = {layer.name: layer for layer in model.layers}
layer_name = 'conv2d_5'
layer_output = layer_dict[layer_name].output
n_filters = int(layer_output.shape[-1])
DATA_DIR = os.path.abspath('data')
LOG_FILE = 'driving_log.csv'
# Load the image_data from the driving log
image_data = []
with open(os.path.join(DATA_DIR, LOG_FILE)) as handle:
reader = csv.reader(handle)
for row in reader:
image_data.append(row)
# Get a random line from the driving log csv
line = image_data[random.randint(0, len(image_data))]
image_file = os.path.join(DATA_DIR, line[0])
# image_file = "/Users/bowerslm/Dev/sdc/behavioral-cloning/data/IMG/center_2017_04_17_00_39_56_559.jpg"
def normalize(x):
# utility function to normalize a tensor by its L2 norm
return x / (K.sqrt(K.mean(K.square(x))) + 1e-5)
def deprocess_image(x):
# normalize tensor: center on 0., ensure std is 0.1
x -= x.mean()
x /= (x.std() + 1e-5)
x *= 0.1
# clip to [0, 1]
x += 0.5
x = np.clip(x, 0, 1)
# convert to RGB array
x *= 255
if K.image_data_format() == 'channels_first':
x = x.transpose((1, 2, 0))
x = np.clip(x, 0, 255).astype('uint8')
return x
kept_filters = []
for filter_index in range(0, n_filters):
print('Processing filter %d' % filter_index)
start_time = time.time()
# we build a loss function that maximizes the activation
# of the nth filter of the layer considered
if K.image_data_format() == 'channels_first':
loss = K.mean(layer_output[:, filter_index, :, :])
else:
loss = K.mean(layer_output[:, :, :, filter_index])
# we compute the gradient of the input picture wrt this loss
grads = K.gradients(loss, input_img)[0]
# normalization trick: we normalize the gradient
grads = normalize(grads)
# this function returns the loss and grads given the input picture
iterate = K.function([input_img], [loss, grads])
# step size for gradient ascent
step = 1.
# we start from a gray image with some random noise
if K.image_data_format() == 'channels_first':
input_img_data = np.random.random((1, 3, img_height, img_width))
else:
input_img_data = np.random.random((1, img_height, img_width, 3))
input_img_data = (input_img_data - 0.5) * 20 + 128
# input_img_data = imread(image_file).astype(np.float32)
# input_img_data = input_img_data.reshape((-1,) + input_img_data.shape)
# we run gradient ascent for 20 steps
for i in range(20):
loss_value, grads_value = iterate([input_img_data])
input_img_data += grads_value * step
print('Current loss value:', loss_value)
if loss_value <= 0.:
# some filters get stuck to 0, we can skip them
break
# decode the resulting input image
if loss_value > 0:
img = deprocess_image(input_img_data[0])
kept_filters.append((img, loss_value))
end_time = time.time()
print('Filter %d processed in %ds' % (filter_index, end_time - start_time))
n = 3
# the filters that have the highest loss are assumed to be better-looking.
# we will only keep the top nxn filters.
kept_filters.sort(key=lambda x: x[1], reverse=True)
kept_filters = kept_filters[:n * n]
margin = 5
width = n * int(img_width) + (n - 1) * margin
height = n * int(img_height) + (n - 1) * margin
stitched_filters = np.zeros((height, width, 3))
# fill the picture with our saved filters
for i in range(n):
for j in range(n):
img, loss = kept_filters[i * n + j]
stitched_filters[(img_height + margin) * i: (img_height + margin) * i + img_height,
(img_width + margin) * j: (img_width + margin) * j + img_width, :] = img
# save the result to disk
imsave('stitched_filters_%s_%dx%d.png' % (layer_name, n, n), stitched_filters)
print('IMAGE FILE', image_file)