/
visualize.py
202 lines (148 loc) · 7.56 KB
/
visualize.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
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
from googlenet_custom_layers import PoolHelper,LRN
from keras.models import model_from_json
from scipy import ndimage, misc
import cv2
import numpy as np
import os
import pickle
import pandas as pd
from pandas import HDFStore, DataFrame
import h5py
import matplotlib.pyplot as plt
import sys
sys.setrecursionlimit(10000)
from scipy.misc import imread, imresize
from keras.layers import Convolution2D, MaxPooling2D, AveragePooling2D, ZeroPadding2D
from keras.layers import Flatten, merge, Reshape, Activation
from keras.layers import Input, LSTM, Dense, Activation, GRU,Convolution1D,Dropout
from keras.layers.pooling import GlobalAveragePooling2D, GlobalMaxPooling2D,GlobalMaxPooling1D
from keras.models import Model
from keras.regularizers import l2
from keras.optimizers import SGD, RMSprop,Adam
from keras.utils.np_utils import to_categorical
from keras.callbacks import CSVLogger, ReduceLROnPlateau, ModelCheckpoint
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.preprocessing.image import ImageDataGenerator
from googlenet_custom_layers import PoolHelper,LRN
from datetime import datetime
from utils.utils import process_image, deprocess_image
from utils.utils import read_and_generate_heatmap, prepare_data,evaluate_distribution_accuracy
from models import create_model
max_features = 20000
maxlen=100
EMBEDDING_DIM = 300
use_distribution = True
use_semantics = False
use_comments=True
use_multigap=True
# X_train, Y_train, X_test, Y_test = prepare_data(use_distribution=use_distribution)
X_train, Y_train,X_test, Y_test,X_train_text, X_test_text,embedding_layer =
prepare_data(use_distribution=use_distribution, use_semantics=use_semantics, use_comments=use_comments)
# X_train, Y_train,X_test, Y_test= prepare_data(use_distribution=use_distribution, use_semantics=False)
# X_train, Y_train, Y_train_semantics, X_test, Y_test, Y_test_semantics, X_train_text, X_test_text, embedding_layer = prepare_data(use_distribution=use_distribution, use_semantics=use_semantics, use_comments=use_comments)
## Without image data
# _, Y_train,_, Y_test,X_train_text, X_test_text,embedding_layer = prepare_data(use_distribution=use_distribution, use_semantics=use_semantics, use_comments=use_comments, imageDataAvailable=False)
# BEST MODEL
model = create_model('weights/2017-01-25 22_56_09 - distribution_2layergru_extra_conv_layer.h5',
use_distribution=use_distribution, use_semantics=use_semantics,use_multigap=use_multigap,use_comments=use_comments,
embedding_layer=embedding_layer,extra_conv_layer=True,textInputMaxLength=maxlen,embedding_dim=EMBEDDING_DIM)
# model = create_model('weights/googlenet_aesthetics_weights.h5',
# use_distribution=use_distribution, use_semantics=use_semantics,use_multigap=True, heatmap=False)
# MODEL WITH EXTRA CONV AND NO TEXT
# model = create_model('weights/2017-01-27 12:41:36 - distribution_extra_conv_layer.h5',
# use_distribution=use_distribution, use_semantics=use_semantics,
# use_multigap=True,extra_conv_layer=True)
# RAPID STYLE
# model = create_model('weights/googlenet_aesthetics_weights.h5',
# use_distribution=use_distribution, use_semantics=True,
# use_multigap=False,extra_conv_layer=False, rapid_style=True)
# rmsProp = RMSprop(lr=0.0001,clipnorm=1.,clipvalue=0.5)
adam = Adam(lr=0.0001,clipnorm=1.,clipvalue=0.5)
if use_distribution:
print("using kld loss...")
model.compile(optimizer=adam,loss='kld', metrics=['accuracy'])#,loss_weights=[1., 0.2])
else:
print("using categorical crossentropy loss...")
model.compile(optimizer=adam,loss='categorical_crossentropy', metrics=['accuracy'])#,loss_weights=[1., 0.2])
from keras.utils.visualize_util import plot
plot(model, to_file='{}.png'.format(unique_model_identifier),show_shapes=True)
from keras import backend as K
def get_output_layer(model, layer_name):
# get the symbolic outputs of each "key" layer (we gave them unique names).
layer_dict = dict([(layer.name, layer) for layer in model.layers])
layer = layer_dict[layer_name]
return layer
gap_conv_layer_4a = get_output_layer(model, "conv_4a")
gap_conv_layer_4b = get_output_layer(model, "conv_4b")
gap_conv_layer_4c = get_output_layer(model, "conv_4c")
gap_conv_layer_4d = get_output_layer(model, "conv_4d")
final_conv_layer = get_output_layer(model, "conv_6_1")
if use_multigap:
if use_comments:
get_output = K.function(
[ model.inputs[0], model.inputs[1],K.learning_phase() ] ,
[final_conv_layer.output,gap_conv_layer_4a.output,
gap_conv_layer_4b.output,gap_conv_layer_4c.output,
gap_conv_layer_4d.output, model.layers[-1].output])
else:
get_output = K.function(
[ model.inputs[0],K.learning_phase() ] ,
[final_conv_layer.output,gap_conv_layer_4a.output,
gap_conv_layer_4b.output,gap_conv_layer_4c.output,
gap_conv_layer_4d.output, model.layers[-1].output])
else:
get_output = K.function(
[ model.inputs[0],K.learning_phase() ] ,
[final_conv_layer.output, model.layers[-1].output])
class_weights = model.layers[-1].get_weights()[0]
images_to_show = 50
### To show a range of values
#
# total_amount = X_test_text.shape[0]
# middle = int(total_amount/2)
# min_boundary = middle - int(images_to_show/2)
# max_boundary = middle + int(images_to_show/2)
# X_test_text_used = X_test_text[min_boundary:max_boundary]#[::-1]
#
###
class_weights_to_visualize = class_weights[1324:1948]
class_weights_to_visualize = np.column_stack((
class_weights_to_visualize[:,0:5].mean(axis=1),
class_weights_to_visualize[:,5:10].mean(axis=1)))
X_test_text_used = X_test_text[-images_to_show:][::-1]
for comment_idx, index in enumerate(ava_test[-images_to_show:][::-1].index):
output_filename = "heatmaps/{} - comments.png".format(index)
if os.path.isfile(output_filename):
print("[INFO] file with id of {} already exists, skipping".format(index))
else:
input_path = "datasetdataset/AVA/data/{}.jpg".format(index)
original_img = cv2.imread(input_path).astype(np.float32)
width, height,_ = original_img.shape
original_img = cv2.resize(original_img, (int(height / 2),int(width /2)))
width, height,_ = original_img.shape
im = process_image(cv2.resize(original_img,(224,224)))
[conv_outputs, gap_conv_outputs_4a,gap_conv_outputs_4b,
gap_conv_outputs_4c,gap_conv_outputs_4d, predictions] = get_output( [im,
np.expand_dims(X_test_text_used[comment_idx], axis=0),
0])
# [conv_outputs, predictions] = get_output( [im,0])
conv_to_visualize = gap_conv_outputs_4a[0, :, :, :]
# conv_to_visualize = conv_outputs[0, :, :, :]
output_image = original_img.copy()
for class_weight_to_visualize in class_weights_to_visualize.T:
cam = np.zeros(dtype = np.float32, shape = conv_to_visualize.shape[1:3])
class_to_visualize = 1 # 0 for bad, 1 for good
for i, w in enumerate(class_weight_to_visualize):
cam += w * conv_to_visualize[i, :, :]
cam /= np.max(cam)
cam = cv2.resize(cam, (height, width))
heatmap = cv2.applyColorMap(np.uint8(255*cam), cv2.COLORMAP_JET)
heatmap[np.where(cam < 0.2)] = 0
img_cam = heatmap*0.5 + original_img
print("CALLED CONCATENATE")
output_image = np.concatenate((output_image, img_cam), axis=1)
# output_image = cv2.resize(output_image, (int(output_image.shape[1] / 2), int(output_image.shape[0]/2)))
cv2.imwrite(output_filename, output_image)
print()
print()