-
Notifications
You must be signed in to change notification settings - Fork 2
/
Keras_CNN2.py
192 lines (163 loc) · 6.59 KB
/
Keras_CNN2.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
import numpy as np
from OriginalImg import OriginalImg
import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
from keras.optimizers import SGD
from keras.utils import np_utils
from keras.regularizers import activity_l2
import os
from project_types import *
import pickle
# from keras.layers import Dense, Activation
class Keras_CNN():
def __init__(self):
self.model = Sequential()
self.model.add(Convolution2D(32, 3, 3, border_mode='same',
input_shape=(128,128, 3)))
self.model.add(Activation('relu'))
self.model.add(Convolution2D(32, 3, 3))
self.model.add(Activation('relu'))
self.model.add(MaxPooling2D(pool_size=(2, 2)))
# self.model.add(Convolution2D(32, 3, 3))
self.model.add(Dropout(0.25))
self.model.add(Convolution2D(64, 3, 3, border_mode='same'))
self.model.add(Activation('relu'))
self.model.add(Convolution2D(64, 3, 3))
self.model.add(Activation('relu'))
self.model.add(MaxPooling2D(pool_size=(2, 2)))
self.model.add(Dropout(0.25))
self.model.add(Flatten())
self.model.add(Dense(128))
self.model.add(Activation('relu'))
# self.model.add(Dropout(0.5))
self.model.add(Dense(6))
# self.model.add(Activation('sigmoid'))
# let's train the model using SGD + momentum (how original).
sgd = SGD(lr=1e-6, decay=1e-6, momentum=0.9, nesterov=True)
self.model.compile(loss='hinge',
optimizer=sgd,
metrics=['accuracy'])
# Y_train = np_utils.to_categorical([1,2], 3)
# Y_train = np.array([1,2])
# X_train = np.array([[[[1]], [[2]], [[3]]], [[[6]], [[-1]], [[4]]]])
# print X_train.shape
# self.model.fit(x = X_train, y = Y_train)
def add_convolutional(filters, con_size=(3,3), input_size=None):
# self.model.add()
# apply a 3x3 convolution with 64 output filters on a 256x256 image:
# model = Sequential()
# size should be (3, row, col) for RGB img
if input_size is not None:
self.model.add(keras.layers.convolutional.Convolution2D(filters, con_size[0], con_size[1], border_mode='same', input_shape=(input_size)))
else:
# add a 3x3 convolution on top, with 32 output filters:
model.add(keras.layers.convolutional.Convolution2D(filters, con_size[0], con_size[1], border_mode='same'))
def add_pool(self,size=(2,2)):
self.model.add(keras.layers.pooling.MaxPooling2D(pool_size=size, strides=None, border_mode='valid', dim_ordering='default'))
def add_relu(self):
self.model.add(Activation('relu'))
def compile_model(self):
self.model.compile(optimizer='rmsprop',
loss='categorical_crossentropy',
metrics=['accuracy'])
def train(self, feature_converter):
self.labels = set()
for i in range(feature_converter.trainingInstancesSize()):
label = feature_converter.getTrainingLabel(i)
self.labels.add(str(label))
self.labels = list(self.labels)
self.label_dict = {}
i = 0
for label in self.labels:
self.label_dict[str(label)] = i
i += 1
x = []
y = []
print feature_converter.trainingInstancesSize()
for i in range(feature_converter.trainingInstancesSize()):
inst = feature_converter.getTrainingInstance(i)
vector = inst.get_vector()
# vector = np.swapaxes(vector, 2, 1)
# vector = np.swapaxes(vector, 0, 1)
x.append(vector)
y.append(self.label_dict[inst.get_label()])
# break
x = np.array(x)
print x.shape
y = np.array(y)
y = np_utils.to_categorical(y, 6)
self.model.fit(x, y, nb_epoch=10, batch_size = 32)
def predict(self, feature_converter):
labels = []
total = 0
correct = 0
x = []
y = []
print self.labels[4]
for i in range(feature_converter.testingInstancesSize()):
instance = feature_converter.getTestingInstance(i)
vector = instance.get_vector()
# vector = np.swapaxes(vector, 2, 1)
# vector = np.swapaxes(vector, 0, 1)
x.append(vector)
y.append(self.label_dict[instance.get_label()])
# x = instance.get_vector()
# lab = self.model.predict(x)
# print self.labels[lab], str(instance.get_label())
# # labels.append(self.labels[max_index])
# if self.labels[lab] == str(instance.get_label()):
# correct += 1
# total += 1
x = np.array(x)
y = np.array(y)
z = np_utils.to_categorical(y, 6)
score = self.model.evaluate(x, z)
# print self.model.predict_classes(x)
# print self.label_dict
# print score
# labs = self.model.predict(x)
# count = 0
# correct = 0
# for row in labs:
# max_val = None
# max_index = None
# for i in range(len(row)):
# if max_index is None or row[i] > max_val:
# max_val = row[i]
# max_index = i
# if self.labels[max_index] == self.labels[y[count]]:
# correct += 1
# count += 1
# # # print labs
# print "Testing Accuracy: ", (correct / float(count) * 100), "%"
return labels
def get_files(folder_name):
foldersTmp = os.listdir(folder_name)
folders = []
for folder in foldersTmp:
if folder[0] == '.':
continue
folders.append(folder)
imgs = []
for folder in folders:
path = folder_name + folder + '/'
if not os.path.isdir(path):
continue
files = os.listdir(path)
for file_str in files:
complete_file_str = str((os.path.join(path, file_str)))
if os.path.isfile(complete_file_str) and (complete_file_str.endswith('.jpg') or complete_file_str.endswith('.JPG')):
imgs.append((os.path.join(path, file_str), folder))
return imgs
feature_converter = OriginalImg()
training_files = get_files('data/train/')
testing_files = get_files('data/test/')
feature_converter.createTrainingInstances(training_files)
feature_converter.createTestingInstances(testing_files)
# with open('')
for _ in range(1):
cnn = Keras_CNN()
cnn.train(feature_converter)
cnn.predict(feature_converter)