-
Notifications
You must be signed in to change notification settings - Fork 20
/
dataset.py
163 lines (133 loc) · 6.71 KB
/
dataset.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
"""
Copyright (C) 2018 Axel Davy
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
"""
import os
import os.path
import numpy as np
import numpy.random
import cv2
import torch
import torch.utils.data as udata
import random
from video_patch_search import VideoPatchSearch
# Using the same weights for each channel
# gives optimal contrast over noise.
def rgb_to_gray(img):
img = np.asarray(img, dtype=np.float32)
if (len(img.shape) == 3 or img.shape[3] == 1):
return img
res = np.dot(img[...,:3], [0.57735, 0.57735, 0.57735])
return np.asarray(res, dtype=np.float32)
class Dataset(udata.Dataset):
def __init__(self, data_path, color_mode=False, sigma=25, oracle_mode=0, past_frames=7, future_frames=7, search_window_width=41, nn_patch_width=41, pass_nn_value=False, patch_width=44, patch_stride=5):
super(Dataset, self).__init__()
self.color = color_mode
self.sigma = sigma
categories = os.scandir(data_path)
categories = [c for c in categories if not c.name.startswith('.') and c.is_dir()]
video_paths_dict = {}
video_paths_list = []
for c in categories:
list_for_category = []
paths = os.scandir(c.path)
paths = [p for p in paths if not p.name.startswith('.') and p.is_dir()]
video_paths_list.extend(paths)
video_paths_dict[c.name] = [p.path for p in paths]
categories = [c.name for c in categories]
print ('%d categories' % len(categories))
print ('%d videos' % len(video_paths_list))
self.categories = categories
self.video_paths_dict = video_paths_dict
self.patch_width_nn = nn_patch_width
self.patch_data_width_nn = 1
self.past_frames = past_frames
self.future_frames = future_frames
self.num_neighbors = 1 + past_frames + future_frames
self.patch_width = patch_width
self.patch_stride = patch_stride
self.oracle_mode = oracle_mode
self.pass_nn_value = pass_nn_value
self.ps = VideoPatchSearch(patch_search_width=self.patch_width_nn, patch_data_width=self.patch_data_width_nn,
past_frames=self.past_frames, future_frames=self.future_frames,
input_dtype=np.float32,
search_width=search_window_width)
np.random.seed(2018)
self.prepare_epoch()
def prepare_epoch(self):
"""
Read random videos of the database
"""
subdirnames = ['/01/', '/02/', '/03/', '/04/', '/05/', '/06/', '/07/', '/08/', '/09/']
filenames= ['001.png', '002.png', '003.png', '004.png', '005.png', '006.png', '007.png', '008.png', '009.png', '010.png', '011.png', '012.png', '013.png', '014.png', '015.png']
burst_nums = np.random.randint(len(subdirnames), size=len(self.categories*10))
frame_nums = np.random.randint(self.past_frames, high=len(filenames)-self.future_frames, size=len(self.categories*10))
i = 0
self.videos = []
self.keys = []
for c in self.categories:
paths = self.video_paths_dict[c]
paths = np.random.permutation(paths)
for p in paths[:3]:
dir_path = p + subdirnames[burst_nums[i]]
if not (os.path.exists(dir_path)):
continue
video = []
for f in range(frame_nums[i]-self.past_frames, frame_nums[i]+self.future_frames+1):
img = cv2.imread(dir_path + filenames[f])
img = np.asarray(img, dtype=np.float32)
if self.color:
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
else:
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img = np.expand_dims(img, 2)
img = np.asarray(img, dtype=np.float32)
img = img/255.
video.append(img)
ref_image = video[self.past_frames]
video = np.asarray(video, dtype=np.float32)
video_noised = video + self.sigma * np.random.randn(video.shape[0], video.shape[1], video.shape[2], video.shape[3])
video_noised = np.asarray(video_noised, dtype=np.float32)
video_search_gray = rgb_to_gray(video if self.oracle_mode == 1 else video_noised)
nn = self.ps.compute(video_search_gray, self.past_frames)
self.videos.append((video_noised[self.past_frames, ...] - ref_image, video_noised, nn))
ys = range(2*self.patch_width, ref_image.shape[0]-2*self.patch_width, self.patch_stride)
xs = range(2*self.patch_width, ref_image.shape[1]-2*self.patch_width, self.patch_stride)
xx, yy = np.meshgrid(xs, ys)
xx = np.asarray(xx.flatten(), dtype=np.uint32)
yy = np.asarray(yy.flatten(), dtype=np.uint32)
self.keys.append(np.stack([i*np.ones([len(xx)], dtype=np.uint32), xx, yy]).T)
i = i + 1
self.keys = np.concatenate(self.keys, axis=0)
self.num_keys = self.keys.shape[0]
self.indices = [i for i in range(self.num_keys)]
random.shuffle(self.indices)
def __len__(self):
return self.num_keys
def data_num_channels(self):
return (3 if self.color else 1) * (self.num_neighbors if self.pass_nn_value else 1)
def __getitem__(self, index):
key = self.keys[self.indices[index],:]
patch_width = self.patch_width
i = key[0]
x = key[1]
y = key[2]
anchor = self.patch_width_nn//2
noise = self.videos[i][0][(y-patch_width):y, (x-patch_width):x,:]
if self.pass_nn_value:
nn_patch = self.videos[i][2][(y-anchor-patch_width):(y-anchor), (x-anchor-patch_width):(x-anchor),:]
patch_stack = self.ps.build_neighbors_array(self.videos[i][1], nn_patch)
else:
patch_stack = self.videos[i][1][self.past_frames, (y-patch_width):y, (x-patch_width):x,:]
patch_stack = patch_stack.transpose(2, 0, 1)
noise = noise.transpose(2, 0, 1)
return (torch.Tensor(patch_stack), torch.Tensor(noise))