-
Notifications
You must be signed in to change notification settings - Fork 111
/
datasets.py
executable file
·201 lines (142 loc) · 6.49 KB
/
datasets.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
import numpy as np
from torch.utils.data import Dataset
import os
from glob import glob
from PIL import Image
from utils.file_io import read_img
class ScannetDataset(Dataset):
def __init__(self,
data_dir='datasets/Scannet',
mode='train',
transforms=None,
return_filename=False,
):
self.data_dir = data_dir
self.transforms = transforms
self.return_filename = return_filename
assert mode in ['train', 'test', 'demo']
self.mode = mode
self.samples = []
# following BA-Net's splits
dir_path = os.path.dirname(os.path.realpath(__file__))
split_file = 'scannet_banet_' + mode + '_pairs.txt'
split_file = os.path.join(dir_path, split_file)
with open(split_file) as f:
pairs = f.readlines()
pairs = [pair.rstrip() for pair in pairs]
for i in range(len(pairs)):
scene, img_ref_id, img_tgt_id = pairs[i].split(' ')
key = 'scannet_' + mode + '_' + scene + '_' + img_ref_id
scene = os.path.join(data_dir, mode, scene)
intrinsics = os.path.join(scene, 'intrinsic', 'intrinsic_depth.txt')
img_ref = os.path.join(scene, 'color', img_ref_id + '.jpg')
img_tgt = os.path.join(scene, 'color', img_tgt_id + '.jpg')
depth = os.path.join(scene, 'depth', img_ref_id + '.png')
pose_ref = os.path.join(scene, 'pose', img_ref_id + '.txt')
pose_tgt = os.path.join(scene, 'pose', img_tgt_id + '.txt')
if not os.path.isfile(img_ref) or not os.path.isfile(img_tgt) or not os.path.isfile(depth) or \
not os.path.isfile(pose_ref) or not os.path.isfile(pose_tgt):
continue
sample = (img_ref, img_tgt, pose_ref, pose_tgt, depth, intrinsics, key)
self.samples.append(sample)
def __getitem__(self, i):
img_ref, img_tgt, pose_ref, pose_tgt, depth, intrinsics, key = self.samples[i]
img_ref_filename, img_tgt_filename = img_ref, img_tgt
img_ref = self._read_image(img_ref)
img_tgt = self._read_image(img_tgt)
depth = self._read_depth(depth)
valid = (depth > 0).astype(np.float32) # invalid depth is 0
# pose: camera to world
pose_ref = np.loadtxt(pose_ref, delimiter=' ').astype(np.float32).reshape((4, 4))
pose_tgt = np.loadtxt(pose_tgt, delimiter=' ').astype(np.float32).reshape((4, 4))
# relative pose
pose = np.linalg.inv(pose_tgt) @ pose_ref
intrinsics = np.loadtxt(intrinsics).astype(np.float32).reshape((4, 4))[:3, :3] # [3, 3]
sample = {
'img_ref': img_ref,
'img_tgt': img_tgt,
'intrinsics': intrinsics,
'pose': pose,
'depth': depth,
'valid': valid,
}
if self.transforms is not None:
sample = self.transforms(sample)
if self.return_filename:
return img_ref_filename, img_tgt_filename, sample
return sample
def __len__(self):
return len(self.samples)
def _read_image(self, filename):
img = Image.open(filename).resize((640, 480)) # resize to depth shape
img = np.array(img).astype(np.float32)
return img
def _read_depth(self, filename):
depth = np.array(Image.open(filename)).astype(np.float32) / 1000.
return depth
def __rmul__(self, v):
self.samples = v * self.samples
return self
class DemonDataset(Dataset):
def __init__(self,
data_dir='datasets/Demon',
mode='train',
transforms=None,
sequence_length=2,
):
if 'test' in mode:
data_dir = os.path.join(data_dir, 'test')
else:
data_dir = os.path.join(data_dir, 'train')
self.data_dir = data_dir
self.transforms = transforms
assert sequence_length == 2 # only support two input views currently
self.samples = []
scenes = [os.path.join(data_dir, scene_dir) for scene_dir in sorted(os.listdir(data_dir))
if os.path.isdir(os.path.join(os.path.join(data_dir, scene_dir))) and mode in scene_dir]
demi_length = sequence_length // 2
for scene in scenes:
intrinsics = np.genfromtxt(os.path.join(scene, 'cam.txt')).astype(np.float32).reshape((3, 3)) # [3, 3]
poses = np.genfromtxt(os.path.join(scene, 'poses.txt')).astype(np.float32)
imgs = sorted(glob(os.path.join(scene, '*.jpg')))
if len(imgs) < sequence_length:
continue
for i in range(len(imgs)):
if i < demi_length:
shifts = list(range(0, sequence_length))
shifts.pop(i)
elif i >= len(imgs) - demi_length:
shifts = list(range(len(imgs) - sequence_length, len(imgs)))
shifts.pop(i - len(imgs))
else:
shifts = list(range(i - demi_length, i + (sequence_length + 1) // 2))
shifts.pop(demi_length)
img_ref = imgs[i]
depth = os.path.join(os.path.dirname(img_ref), os.path.basename(img_ref)[:-4] + '.npy')
pose_ref = np.concatenate((poses[i, :].reshape((3, 4)), np.array([[0, 0, 0, 1]])), axis=0) # [4, 4]
assert len(shifts) < 2 # only support two input images currently
for j in shifts:
img_tgt = imgs[j]
pose_tgt = np.concatenate((poses[j, :].reshape((3, 4)), np.array([[0, 0, 0, 1]])), axis=0)
pose = (pose_tgt @ np.linalg.inv(pose_ref)).astype(np.float32) # [4, 4]
sample = (img_ref, img_tgt, pose, depth, intrinsics)
self.samples.append(sample)
def __getitem__(self, i):
img_ref, img_tgt, pose, depth, intrinsics = self.samples[i]
img_ref = read_img(img_ref)
img_tgt = read_img(img_tgt)
depth = np.load(depth)
valid = (depth > 0).astype(np.float32) # invalid depth is 0
sample = {
'img_ref': img_ref,
'img_tgt': img_tgt,
'intrinsics': intrinsics,
'pose': pose,
'depth': depth,
'valid': valid,
}
if self.transforms is not None:
sample = self.transforms(sample)
return sample
def __len__(self):
return len(self.samples)