cs231nProject/extract_data.py at master · histopathology/cs231nProject · GitHub

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
from os import listdir
from os.path import isdir, join
from PIL import Image
import numpy as np
import os
import csv

FRAMES_PATH = "/frames/x40/"
ATYPIA_PATH = "/atypia/x40/"
SIZE = 36,36

def imageToVector(directory):
	frames = directory + FRAMES_PATH
	atypia = directory + ATYPIA_PATH
	frames_files = [join(frames,f) for f in listdir(frames)]
	atypia_files = [join(atypia, f) for f in listdir(atypia)]
	num_files = len(frames_files)
	data = []
	scores = []

	for i in range(num_files):
		with open(atypia_files[i], 'r') as f:
			lines = list(csv.reader(f, delimiter=','))
			"""if len(lines) > 0 and int(lines[0][0]) < 3:
				scores.append(int(lines[0][0]))
				img = Image.open(frames_files[i])
				img.thumbnail(SIZE, Image.ANTIALIAS)
				arr = np.array(img)
				arr = arr[:, 0:arr.shape[0],:]
				data.append(arr)"""
			if len(lines[0]) > 1:
				score = int(max(set(lines[1][1:4]), key=lines[1][1:4].count))
				if score < 3:
					scores.append(int(max(set(lines[1][1:4]), key=lines[1][1:4].count)))
					img = Image.open(frames_files[i])
					img.thumbnail(SIZE, Image.ANTIALIAS)
					arr = np.array(img)
					arr = arr[:, 0:arr.shape[0],:]
					data.append(arr)

	return np.array(data), np.array(scores)

def main():
	directories = [f for f in listdir(os.getcwd()) if isdir(f) and 'git' not in f]
	X_a = []
	y_a = []
	X_h = []
	y_h = []
	X = []
	y = []

	# Extract data from files into X and y
	for directory in directories:
		data, scores = imageToVector(directory)
		a,b,c,d = data.shape
		data = data.reshape(a,d,b,c)
		X.append(data)
		y.append(scores)
		if "A" in directory:
			X_a.append(data)
			y_a.append(scores)
		else:
			X_h.append(data)
			y_h.append(scores)
	X = np.concatenate(np.array(X), axis=0) / 255.0
	y = np.concatenate(np.array(y), axis=0) - 1
	X_a = np.concatenate(np.array(X_a), axis=0) / 255.0
	y_a = np.concatenate(np.array(y_a), axis=0) - 1
	X_h = np.concatenate(np.array(X_h), axis=0) / 255.0
	y_h = np.concatenate(np.array(y_h), axis=0) - 1
	print(np.count_nonzero(y_a == 0))
	print(np.count_nonzero(y_a == 1))
	length = X_a.shape[0]
	print(X_a[0].shape)
	print(len(X_a), len(X_h))

	# Randomize
	index = np.random.permutation(length)
	index_full = np.random.permutation(2 * length)
	X_a= X_a[index]
	y_a = y_a[index]
	X_h = X_h[index]
	y_h = y_h[index]
	X = X[index_full]
	y = y[index_full]

	#Split into train and val sets
	cutoff = int(length * .7)
	X_a_train = X_a[0:cutoff,:,:,:]
	y_a_train = y_a[0:cutoff]
	X_a_val = X_a[cutoff:length,:,:,:]
	y_a_val = y_a[cutoff:length]
	X_h_train = X_h[0:cutoff,:,:,:]
	y_h_train = y_h[0:cutoff]
	X_h_val = X_h[cutoff:length,:,:,:]
	y_h_val = y_h[cutoff:length]
	X_train = X[0:2*cutoff,:,:,:]
	y_train = y[0:2*cutoff]
	X_val = X[2*cutoff:2*length,:,:,:]
	y_val = y[2*cutoff:2*length]
	print(len(X_train), len(X_val))


	#Save files
	np.save('X_a_train.npy', X_a_train)
	np.save('y_a_train.npy', y_a_train)
	np.save('X_a_val.npy', X_a_val)
	np.save('y_a_val.npy', y_a_val)
	np.save('X_h_train.npy', X_h_train)
	np.save('y_h_train.npy', y_h_train)
	np.save('X_h_val.npy', X_h_val)
	np.save('y_h_val.npy', y_h_val)
	np.save('X_train.npy', X_train)
	np.save('y_train.npy', y_train)
	np.save('X_val.npy', X_val)
	np.save('y_val.npy', y_val)

if __name__ == "__main__":
	main()