-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcnn_genetic.py
118 lines (107 loc) · 6.16 KB
/
cnn_genetic.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
# -*- coding: utf-8 -*-
"""
Created on Sun Mar 25 07:27:22 2018
Uses a genetic algorthim to evolve a combination of morphological operators to apply the CIFAR-10
dataset as a function of a CNN's accuracy (trained on the morphed data).
@author: Etcyl
"""
from keras.datasets import cifar10
import build_cnn
import genetic
import keras
import cv2
import numpy as np
#Set constants here
batch_size = 32
num_classes = 10
epochs = 1#100
data_augmentation = True
num_predictions = 20
kernel = np.ones((3,3),np.uint8)
#Create a genetic algorithm class to use the genetic algorithm
evolver = genetic.morphological_evolver()
for k in range(evolver.generations):
#Use each chromosome in the population to train the CNN
for index in range(evolver.pop_size):
#Get the CIFAR-10 data and training sets
(x_train, y_train), (x_test, y_test) = cifar10.load_data()
#Convert class vectors to binary class matrices
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255
#Create the CNN class using Keras
cnn = build_cnn.buildCNN()
if(k == 0): #Update every individual for the first generation, k == 0, and then only the children after
#If a gene is present, decode and apply it to the dataset before training the CNN
for j in range(evolver.num_genes):
if evolver.current_pop[index].getGene(j) == 1:
if j == 1: #Erosion
for i in range(len(x_train)):
x_train[i] = cv2.erode(x_train[i], kernel, iterations = 1)
for i in range(len(x_test)):
x_test[i] = cv2.erode(x_test[i], kernel, iterations = 1)
elif j == 2: #Dilation
for i in range(len(x_train)):
x_train[i] = cv2.dilate(x_train[i], kernel, iterations = 1)
for i in range(len(x_test)):
x_test[i] = cv2.dilate(x_test[i], kernel, iterations = 1)
elif j == 3: #Opening
for i in range(len(x_train)):
x_train[i] = cv2.morphologyEx(x_train[i], cv2.MORPH_OPEN, kernel)
for i in range(len(x_test)):
x_test[i] = cv2.morphologyEx(x_test[i], cv2.MORPH_OPEN, kernel)
elif j == 4: #Closing
for i in range(len(x_train)):
x_train[i] = cv2.morphologyEx(x_train[i], cv2.MORPH_CLOSE, kernel)
for i in range(len(x_test)):
x_test[i] = cv2.morphologyEx(x_test[i], cv2.MORPH_CLOSE, kernel)
elif j == 5: #Gradient
for i in range(len(x_train)):
x_train[i] = cv2.morphologyEx(x_train[i], cv2.MORPH_GRADIENT, kernel)
for i in range(len(x_test)):
x_test[i] = cv2.morphologyEx(x_test[i], cv2.MORPH_GRADIENT, kernel)
cnn.fit(x_train[0:1000], y_train[0:1000], batch_size=batch_size, epochs=epochs,
validation_data=(x_test[0:1000], y_test[0:1000]), shuffle=True)
scores = cnn.evaluate(x_test[0:1000], y_test[0:1000], verbose=1)
#print('Test loss:', scores[0])
print('Test accuracy:', scores[1])
evolver.current_pop[index].setAccuracy(scores[1])
elif(k > 0): #First generation is complete, so only update for future children
if(evolver.current_pop[index].getAccuracy() == 0): #Find children by getting individuals that have 0 accuracy
for j in range(evolver.num_genes):
if evolver.current_pop[index].getGene(j) == 1:
if j == 0: #Erosion
for i in range(len(x_train)):
x_train[i] = cv2.erode(x_train[i], kernel, iterations = 1)
for i in range(len(x_test)):
x_test[i] = cv2.erode(x_test[i], kernel, iterations = 1)
elif j == 1: #Dilation
for i in range(len(x_train)):
x_train[i] = cv2.dilate(x_train[i], kernel, iterations = 1)
for i in range(len(x_test)):
x_test[i] = cv2.dilate(x_test[i], kernel, iterations = 1)
elif j == 2: #Opening
for i in range(len(x_train)):
x_train[i] = cv2.morphologyEx(x_train[i], cv2.MORPH_OPEN, kernel)
for i in range(len(x_test)):
x_test[i] = cv2.morphologyEx(x_test[i], cv2.MORPH_OPEN, kernel)
elif j == 3: #Closing
for i in range(len(x_train)):
x_train[i] = cv2.morphologyEx(x_train[i], cv2.MORPH_CLOSE, kernel)
for i in range(len(x_test)):
x_test[i] = cv2.morphologyEx(x_test[i], cv2.MORPH_CLOSE, kernel)
elif j == 4: #Gradient
for i in range(len(x_train)):
x_train[i] = cv2.morphologyEx(x_train[i], cv2.MORPH_GRADIENT, kernel)
for i in range(len(x_test)):
x_test[i] = cv2.morphologyEx(x_test[i], cv2.MORPH_GRADIENT, kernel)
cnn.fit(x_train[0:1000], y_train[0:1000], batch_size=batch_size, epochs=epochs,
validation_data=(x_test[0:1000], y_test[0:1000]), shuffle=True)
scores = cnn.evaluate(x_test[0:1000], y_test[0:1000], verbose=1)
#print('Test loss:', scores[0])
print('Test accuracy:', scores[1])
evolver.current_pop[index].setAccuracy(scores[1])
evolver.updatePop() #Create a new generation using crossover and mutation