forked from cmusatyalab/change-detection
-
Notifications
You must be signed in to change notification settings - Fork 0
/
places2_dataset.py
176 lines (148 loc) · 6.67 KB
/
places2_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
164
165
166
167
168
169
170
171
172
173
174
175
176
import torch
from torch.utils.data import Dataset
from torchvision import transforms, datasets
from PIL import Image
import numpy as np
from skimage import io
import argparse
import matplotlib.pyplot as plt
import pickle
import time
import h5py
import glob
import os
import utils
import pdb
#from pdb import set_trace as st
class Places2DatasetCityScapesMasks(Dataset):
''''
Dataset for Training Background Inpainting network
'''
def __init__(self, places2_filelist_path, places2_basepath, root='./cityscapes', split="train", training=False, probability=0.5, image_size = (256,256)):
self.root=root
self.split=split
self.files = {}
self.image_size = image_size
with open(places2_filelist_path, 'r') as f:
self.places2_filelist = f.read().splitlines()
self.places2_basepath = places2_basepath
# Cityscapes
self.images_base = os.path.join(self.root, 'leftImg8bit', self.split)
self.annotations_base = os.path.join(self.root, 'gtFine', self.split)
self.files[split] = self.recursive_glob(rootdir=self.images_base, suffix='.png')
if not self.files[split]:
raise Exception("No files for split=[%s] found in %s" % (split, self.images_base))
print("Found %d %s images" % (len(self.files[split]), split))
self.ignore_index = 0
self.void_classes = [0, 1, 2, 3, 4, 5, 6, 9, 10, 15, 16, 18, 23, -1]
self.valid_classes = [ [12,13], [7], [11,14],[22],[33,24,25,32],[8],[19,20,17],[21],[26,27,28,29,30,31]]
self.class_names = ['wall-fence', 'road', 'building-guardrail', 'terrain', 'bicycle-person-motorcycle-rider',
'sidewalk', 'trafficsign-trafficlight-pole', 'vegetation','car-truck-bus-train']
self.resize_image = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize(self.image_size),
])
self.max_rot = 15
self.max_trans = 0.2
self.max_scale_diff = 0.2
self.max_shear_deg = 10
self.training = training
self.p = probability
# self.normalize_transform = transforms.Normalize(mean=[0.485, 0.456, 0.406],
# std=[0.229, 0.224, 0.225])
self.normalize_transform = transforms.Normalize(mean=[0.5, 0.5, 0.5],
std=[0.5, 0.5, 0.5])
def __len__(self):
return len(self.places2_filelist)
def __getitem__(self, idx):
#pdb.set_trace()
index = np.random.randint(len(self.files[self.split]))
img_path = self.files[self.split][index].rstrip()
lbl_path = os.path.join(self.annotations_base,
img_path.split(os.sep)[-2],
os.path.basename(img_path)[:-15] + 'gtFine_labelIds.png')
instance_path = os.path.join(self.annotations_base,
img_path.split(os.sep)[-2],
os.path.basename(img_path)[:-15] + 'gtFine_instanceIds.png')
_tmp = io.imread(instance_path)
segmentIds = np.unique(_tmp)
instance_id = np.random.randint(len(segmentIds))
mask = _tmp == segmentIds[instance_id]
mask = mask.astype(np.float32)
mask = transforms.ToPILImage()(mask)
mask = transforms.Resize(self.image_size)(mask)
# gt = self.decode_seg(_tmp)
# gt = Image.fromarray(gt)
places2_impath = self.places2_basepath + self.places2_filelist[idx]
im = io.imread(places2_impath)
im = self.resize_image(im).convert('RGB')
#im = torch.from_numpy(io.imread(places2_impath)).permute(2,0,1).float() / 255
if self.training:
# Shear
if (np.random.rand() < self.p):
angle = int( self.max_rot * np.random.rand() )
translation = (self.max_trans * np.random.rand() , self.max_trans * np.random.rand() )
scale = self.max_scale_diff * 2. * (np.random.rand() - 0.5) + 1.
shear = int( self.max_shear_deg * np.random.rand() )
im = transforms.functional.affine(im, angle=angle, translate=translation,
scale=scale, shear=shear, resample=Image.NEAREST, fillcolor=None)
mask = transforms.functional.affine(mask, angle=angle, translate=translation,
scale=scale, shear=shear, resample=Image.NEAREST, fillcolor=None)
if (np.random.rand() < self.p):
im = transforms.functional.hflip(im)
mask = transforms.functional.hflip(mask)
im = self.normalize_transform( transforms.ToTensor()(im).float() )
mask = transforms.ToTensor()(mask)
#pdb.set_trace()
ret = {'im':im,
'mask':mask,
}
return ret
def decode_seg(self, mask):
temp = mask
for _voidc in self.void_classes:
mask[temp == _voidc] = np.int64(self.ignore_index)
for i, classes in enumerate(self.valid_classes):
#pdb.set_trace()
for _validc in classes:
mask[temp == _validc] = np.int64(i+1) #self.class_map[_validc]
return mask
def decode_instance(self, mask):
#pdb.set_trace()
temp = mask
for _voidc in self.void_classes:
mask[temp == _voidc] = np.int64(self.ignore_index)
for i, classes in enumerate(self.valid_classes):
#pdb.set_trace()
for _validc in classes:
mask[temp == _validc] = np.int64(i+1) #self.class_map[_validc]
return mask
def recursive_glob(self, rootdir='.', suffix=''):
return [os.path.join(looproot, filename)
for looproot, _, filenames in os.walk(rootdir)
for filename in filenames if filename.endswith(suffix)]
def parse_args():
parser = argparse.ArgumentParser(description='Pretraining background inpainting')
parser.add_argument(
'--places2filelist',
help='location of places2 filedict',
default='./places365_standard/train.txt',
type=str
)
parser.add_argument(
'--places2_basepath',
help='location of places2 filedict',
default='./places365_standard/',
type=str
)
return parser.parse_args()
def main(args):
ds = Places2DatasetCityScapesMasks(args.places2filelist, args.places2_basepath)
for idx in range(len(ds)):
print(idx)
data=ds[idx]
# im_texture=utils.texture_from_images_and_iuv(data['im'].unsqueeze(0),data['im_iuv'].unsqueeze(0))
# utils.plot_texture_map(im_texture.squeeze(0))
if __name__ == '__main__':
args = parse_args()
main(args)