DeepLearning/image_patches.py at master · dusano/DeepLearning · 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
# Based on CS294A/CS294W Programming Assignment Starter Code
from numpy  import *
import scipy.io


def getPatches(numPatches, patchSize):

	images = scipy.io.loadmat('IMAGES.mat')['IMAGES']

	patches = zeros((numPatches, patchSize*patchSize))

	numImages = images.shape[2]
	imageIdxs = random.randint(numImages, size=numPatches)
	sortedImageIdxs = argsort(imageIdxs)

	lastImageIdx = -1
	for i in range(numPatches):
		imageIdx = imageIdxs[sortedImageIdxs[i]]
		if lastImageIdx != imageIdx:
			img = images[:,:,imageIdx]
			lastImageIdx = imageIdx

		x = random.randint(img.shape[0] - patchSize)
		y = random.randint(img.shape[1] - patchSize)

		patch = img[x:x+patchSize, y:y+patchSize]

		patches[sortedImageIdxs[i], :] = patch.reshape(1, patchSize*patchSize)

	# Remove DC (mean of images)
	patches = patches - mean(patches)

	# Truncate to +/-3 standard deviations and scale to -1 to 1
	pstd = 3 * std(patches)
	patches = maximum(minimum(patches, pstd), -pstd) / pstd

	# Rescale from [-1,1] to [0.1,0.9]
	patches = (patches + 1) * 0.4 + 0.1

	return patches