-
Notifications
You must be signed in to change notification settings - Fork 1
/
iqaDataset.py
262 lines (184 loc) · 8.26 KB
/
iqaDataset.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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
'''
The copyright in this software is being made available under this Software
Copyright License. This software may be subject to other third party and
contributor rights, including patent rights, and no such rights are
granted under this license.
Copyright (c) 1995 - 2020 Fraunhofer-Gesellschaft zur Förderung der
angewandten Forschung e.V. (Fraunhofer)
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted for purpose of testing the functionalities of
this software provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the names of the copyright holders nor the names of its
contributors may be used to endorse or promote products derived from this
software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, INCLUDING
WITHOUT LIMITATION THE PATENTS OF THE COPYRIGHT HOLDERS AND
CONTRIBUTORS, ARE GRANTED BY THIS SOFTWARE LICENSE. THE
COPYRIGHT HOLDERS AND CONTRIBUTORS PROVIDE NO WARRANTY OF PATENT
NON-INFRINGEMENT WITH RESPECT TO THIS SOFTWARE.
'''
import torch
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from PIL import Image
import os
import numpy as np
import random
from sklearn.feature_extraction.image import extract_patches_2d
import matplotlib.pyplot as plt
class IqaDataset(Dataset):
def __init__(self, iqaDataFrame, transform=None, return_numpy=True):
"""
Dataset class that wraps around an iqaDataFrame and can be used with
an pytorch dataloader.
Parameters:
-----------
iqaDataFrame : pandas.DataFrame,
DataFrame as produced by iqaDataFrame that has the following
columns: "path_ref", "path_dist", "q_norm", "quality", "dataset",
"refname", "distortion", "height", "width", "fps", "mse"
transform : pytorch transform,
transform to be applied to each sample
return_numpy : bool
If True, images are returned as numpy arrays. If False images are
returned as PIL images.
"""
self.iqaDataFrame = iqaDataFrame
self.transform = transform
self.return_numpy = return_numpy
def __len__(self):
return self.iqaDataFrame.shape[0]
def __getitem__(self, idx):
"""
Returns:
--------
sample: dict
Dictonary holding reference and distorted image as well as meta
information.
"""
# load images
img_ref = Image.open(
self.iqaDataFrame.iloc[idx].path_ref).convert("L")
img_dist = Image.open(
self.iqaDataFrame.iloc[idx].path_dist).convert("L")
if self.return_numpy:
# convert PIL -> numpy.ndarray
img_ref = np.array(img_ref)
img_dist = np.array(img_dist)
h, w = img_ref.shape
img_ref = np.reshape(img_ref, [1, h, w])
img_dist = np.reshape(img_dist, [1, h, w])
# sample is a dict
sample = {"path_ref" : self.iqaDataFrame.iloc[idx].path_ref,
"path_dist" : self.iqaDataFrame.iloc[idx].path_dist,
"img_ref" : img_ref,
"img_dist" : img_dist,
"q_norm" : torch.Tensor(
[self.iqaDataFrame.iloc[idx].q_norm]),
"quality" : torch.Tensor(
[self.iqaDataFrame.iloc[idx].quality]),
"dataset" : self.iqaDataFrame.iloc[idx].dataset,
"refname" : self.iqaDataFrame.iloc[idx].refname,
"distortion" : self.iqaDataFrame.iloc[idx].distortion,
"height" : self.iqaDataFrame.iloc[idx].height,
"width" : self.iqaDataFrame.iloc[idx].width,
"fps" : self.iqaDataFrame.iloc[idx].fps,
"mse" : torch.Tensor([self.iqaDataFrame.iloc[idx].mse])
}
if self.transform:
# apply transforms
sample = self.transform(sample)
return sample
class RandomBlocks(object):
def __init__(self, block_size, max_blocks):
"""
Extract co-located blocks from random spatial positions.
Parameters:
-----------
block_size : int,
Height and width of blocks.
max_blocks : int,
Number of blocks to be extracted.
"""
self.block_size = block_size
self.max_blocks = max_blocks
def __call__(self, sample):
# channels, height, width
c, h, w = sample["img_ref"].shape
# channels_ref + channels_dist, height, width
ref_dist = np.concatenate(
[sample["img_ref"], sample["img_dist"]], axis=0)
# height, width, channels_ref + channels_dist
ref_dist = np.transpose(ref_dist, [1,2,0])
# blocks, height, width, channels_ref + channels_dist
blocks = extract_patches_2d(
ref_dist, patch_size=(self.block_size, self.block_size),
max_patches=self.max_blocks)
# blocks, channels_ref + channels_dist, height, width
blocks = np.transpose(blocks, [0,3,1,2])
# blocks, channels, height, width
sample["img_ref"] = blocks[:, :c]
sample["img_dist"] = blocks[:, c:]
return sample
class ToTensor(object):
"""
The build-in ToTensor class reduces the pixel range
from [0, 255] to [0, 1]. This class keeps the original range.
"""
def __call__(self, sample):
sample["img_ref"] = torch.Tensor(sample["img_ref"])
sample["img_dist"] = torch.Tensor(sample["img_dist"])
return sample
class FlipHorizontal(object):
"""Flip images horizontally with a probability of 0.5."""
def __call__(self, sample):
"""
Parameters:
-----------
sample : dict,
Dictionary holding reference and distorted images. Importantly,
images must be tensors not PIL images,
"""
if random.getrandbits(1):
# axis -2 is width dimension for both B, H, W, C and H, W, C
# formats
sample["img_ref"] = torch.flip(sample["img_ref"], dims=[-2])
sample["img_dist"] = torch.flip(sample["img_dist"], dims=[-2])
return sample
if __name__ == "__main__":
from iqaDataFrame import IqaDataFrame
# runnung as main is useful for debugging
datadir = "/home/becker/repositories/iqe/datasets"
iqedata = IqaDataFrame(
datadir=datadir, datasets=["liveiqa"], nTrain=17, nVal=6)
# the order of the transforms matters as FlipHorizontalTransform expects
# pytorch tensors whereas the other transforms expect numpy arrays
dataset = IqaDataset(
iqedata.df_train, datadir=datadir, transform=transforms.Compose(
[RandomBlocks(320, 1), ToTensor(), FlipHorizontalTransform()]))
dataloader = DataLoader(dataset, batch_size=4)
for i_batch, sample_batched in enumerate(dataloader):
fig, ax = plt.subplots(1, 2, figsize=(16, 8))
ax[0].imshow(sample_batched["img_ref"][0][0].squeeze(), "gray")
ax[1].imshow(sample_batched["img_dist"][0][0].squeeze(), "gray")
plt.show()
print("done")