-
Notifications
You must be signed in to change notification settings - Fork 0
/
data_loader.py
147 lines (119 loc) · 5.07 KB
/
data_loader.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
import os
import pickle
import numpy as np
import matplotlib.pyplot as plt
class DataLoader:
"""Data Loader class. As a simple case, the model is tried on TinyImageNet. For larger datasets,
you may need to adapt this class to use the Tensorflow Dataset API"""
def __init__(self, batch_size, shuffle=False):
self.X_train = None
self.y_train = None
self.img_mean = None
self.train_data_len = 0
self.X_val = None
self.y_val = None
self.val_data_len = 0
self.X_test = None
self.y_test = None
self.test_data_len = 0
self.shuffle = shuffle
self.batch_size = batch_size
def _unpickle(self, filename):
filename = os.path.join("./dataset/cifar-10-batches-py", filename)
with open(filename, 'rb') as file:
data = pickle.load(file, encoding='bytes')
return data
def _load_data(self, filename):
# Load the pickled data-file.
data = self._unpickle(filename)
# Get the raw images.
images = data[b'data']
# Get the class-numbers for each image. Convert to numpy-array.
labels = np.array(data[b'labels'])
return images, labels
def load_data(self):
# train
files = ["data_batch_1", "data_batch_2", "data_batch_3", "data_batch_4", "data_batch_5"]
data = [self._load_data(filename) for filename in files]
images = [images for images, labels in data]
images = np.concatenate(images, axis=0)
images = images.reshape((images.shape[0], 3, 32, 32))
images = images.transpose([0, 2, 3, 1])
from scipy.misc import imresize
images_reshape = np.zeros((images.shape[0], 224, 224, 3))
for i, image in enumerate(images):
images_reshape[i] = imresize(image, (224, 224))
self.X_train = images_reshape
labels = [labels for images, labels in data]
self.y_train = np.concatenate(labels, axis=0)
# test
files = ["test_batch"]
data = [self._load_data(filename) for filename in files]
images = [images for images, labels in data]
images = np.concatenate(images, axis=0)
images = images.reshape((images.shape[0], 3, 32, 32))
images = images.transpose([0, 2, 3, 1])
from scipy.misc import imresize
images_reshape = np.zeros((images.shape[0], 224, 224, 3))
for i, image in enumerate(images):
images_reshape[i] = imresize(image, (224, 224))
self.X_val = images_reshape
labels = [labels for images, labels in data]
self.y_val = np.concatenate(labels, axis=0)
# data statics
self.train_data_len = len(self.X_train)
self.val_data_len = len(self.X_val)
img_height = 224
img_width = 224
num_channels = 3
return img_height, img_width, num_channels, self.train_data_len, self.val_data_len
def generate_batch(self, type='train'):
"""Generate batch from X_train/X_test and y_train/y_test using a python DataGenerator"""
if type == 'train':
# Training time!
new_epoch = True
start_idx = 0
mask = None
while True:
if new_epoch:
start_idx = 0
if self.shuffle:
mask = np.random.choice(self.train_data_len, self.train_data_len, replace=False)
else:
mask = np.arange(self.train_data_len)
new_epoch = False
# Batch mask selection
X_batch = self.X_train[mask[start_idx:start_idx + self.batch_size]]
y_batch = self.y_train[mask[start_idx:start_idx + self.batch_size]]
start_idx += self.batch_size
# Reset everything after the end of an epoch
if start_idx >= self.train_data_len:
new_epoch = True
mask = None
yield X_batch, y_batch
elif type == 'test':
# Testing time!
start_idx = 0
while True:
# Batch mask selection
X_batch = self.X_test[start_idx:start_idx + self.batch_size]
y_batch = self.y_test[start_idx:start_idx + self.batch_size]
start_idx += self.batch_size
# Reset everything
if start_idx >= self.test_data_len:
start_idx = 0
yield X_batch, y_batch
elif type == 'val':
# Testing time!
start_idx = 0
while True:
# Batch mask selection
X_batch = self.X_val[start_idx:start_idx + self.batch_size]
y_batch = self.y_val[start_idx:start_idx + self.batch_size]
start_idx += self.batch_size
# Reset everything
if start_idx >= self.val_data_len:
start_idx = 0
yield X_batch, y_batch
else:
raise ValueError("Please select a type from \'train\', \'val\', or \'test\'")