-
Notifications
You must be signed in to change notification settings - Fork 15
/
data_preprocessing.py
66 lines (56 loc) · 2.36 KB
/
data_preprocessing.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
# python3
# -*- coding: utf-8 -*-
"""
Created on Mon Jul 10 21:22:58 2017
"""
import numpy as np
from scipy.ndimage.interpolation import map_coordinates
from scipy.ndimage.filters import gaussian_filter
def batch_elastic_transform(images, sigma, alpha, height, width, random_state=None):
'''
this code is borrowed from chsasank on GitHubGist
Elastic deformation of images as described in [Simard 2003].
images: a two-dimensional numpy array; we can think of it as a list of flattened images
sigma: the real-valued variance of the gaussian kernel
alpha: a real-value that is multiplied onto the displacement fields
returns: an elastically distorted image of the same shape
'''
assert len(images.shape) == 2
# the two lines below ensure we do not alter the array images
e_images = np.empty_like(images)
e_images[:] = images
e_images = e_images.reshape(-1, height, width)
if random_state is None:
random_state = np.random.RandomState(None)
x, y = np.mgrid[0:height, 0:width]
for i in range(e_images.shape[0]):
dx = gaussian_filter((random_state.rand(height, width) * 2 - 1), sigma, mode='constant') * alpha
dy = gaussian_filter((random_state.rand(height, width) * 2 - 1), sigma, mode='constant') * alpha
indices = x + dx, y + dy
e_images[i] = map_coordinates(e_images[i], indices, order=1)
return e_images.reshape(-1, 784)
if __name__ == "__main__":
import matplotlib.pyplot as plt
from tensorflow.examples.tutorials.mnist import input_data
'''
the following code demonstrates how gaussian_filter works by ploting
the displacement field before and after applying the gaussian_filter
'''
random_state = np.random.RandomState(None)
dx1 = random_state.rand(28, 28) * 2 - 1
dy1 = random_state.rand(28, 28) * 2 - 1
dx2 = gaussian_filter(dx1, 4, mode='constant')
dy2 = gaussian_filter(dy1, 4, mode='constant')
x, y = np.mgrid[0:28, 0:28]
plt.quiver(x, y, dx1, dy1)
plt.show()
plt.quiver(x, y, dx2, dy2)
plt.show()
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)
img = mnist.train.images[0]
plt.imshow(img.reshape(28, -1), cmap='gray')
plt.show()
dimg = elastic_transform(img.reshape(28, -1), sigma=4, alpha=20)
plt.imshow(dimg, cmap='gray')
plt.show()
plt.close()