-
Notifications
You must be signed in to change notification settings - Fork 25
/
data.py
156 lines (138 loc) · 6.05 KB
/
data.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
import os
from PIL import Image
import torch.utils.data as data
import torchvision.transforms as transforms
class SalObjDataset(data.Dataset):
def __init__(self, image_root, gt_root, depth_root, gray_root, trainsize):
self.trainsize = trainsize
self.images = [image_root + f for f in os.listdir(image_root) if f.endswith('.jpg')]
self.gts = [gt_root + f for f in os.listdir(gt_root) if f.endswith('.jpg')
or f.endswith('.png')]
self.depths = [depth_root + f for f in os.listdir(depth_root) if f.endswith('.png')]
self.grays = [gray_root + f for f in os.listdir(gray_root) if f.endswith('.png')]
self.images = sorted(self.images)
self.gts = sorted(self.gts)
self.depths = sorted(self.depths)
self.grays = sorted(self.grays)
self.filter_files()
self.size = len(self.images)
self.img_transform = transforms.Compose([
transforms.Resize((self.trainsize, self.trainsize)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
self.gt_transform = transforms.Compose([
transforms.Resize((self.trainsize, self.trainsize)),
transforms.ToTensor()])
self.depth_transform = transforms.Compose([
transforms.Resize((self.trainsize, self.trainsize)),
transforms.ToTensor()])
self.gray_transform = transforms.Compose([
transforms.Resize((self.trainsize, self.trainsize)),
transforms.ToTensor()])
def __getitem__(self, index):
image = self.rgb_loader(self.images[index])
gt = self.binary_loader(self.gts[index])
depth = self.rgb_loader(self.depths[index])
gray = self.binary_loader(self.grays[index])
image = self.img_transform(image)
gt = self.gt_transform(gt)
depth = self.depth_transform(depth)
gray = self.gray_transform(gray)
# img_names = self.images[index]
return image, gt, depth, gray, index
def filter_files(self):
assert len(self.images) == len(self.gts)
assert len(self.images) == len(self.depths)
assert len(self.images) == len(self.grays)
images = []
gts = []
depths = []
grays = []
for img_path, gt_path, depth_path, gray_path in zip(self.images, self.gts, self.depths, self.grays):
img = Image.open(img_path)
gt = Image.open(gt_path)
depth = Image.open(depth_path)
gray = Image.open(gray_path)
if img.size == gt.size:
images.append(img_path)
gts.append(gt_path)
depths.append(depth_path)
grays.append(gray_path)
self.images = images
self.gts = gts
self.depths = depths
self.grays = grays
def rgb_loader(self, path):
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('RGB')
def binary_loader(self, path):
with open(path, 'rb') as f:
img = Image.open(f)
# return img.convert('1')
return img.convert('L')
def resize(self, img, gt):
assert img.size == gt.size
w, h = img.size
if h < self.trainsize or w < self.trainsize:
h = max(h, self.trainsize)
w = max(w, self.trainsize)
return img.resize((w, h), Image.BILINEAR), gt.resize((w, h), Image.NEAREST)
else:
return img, gt
def __len__(self):
return self.size
def get_loader(image_root, gt_root, depth_root, gray_root, batchsize, trainsize, shuffle=True, num_workers=12, pin_memory=True):
dataset = SalObjDataset(image_root, gt_root, depth_root, gray_root, trainsize)
data_loader = data.DataLoader(dataset=dataset,
batch_size=batchsize,
shuffle=shuffle,
num_workers=num_workers,
pin_memory=pin_memory)
return data_loader, dataset.size
# def update_data_loader(new_z, dataset):
#
# dataset = SalObjDataset(image_root, gt_root, trainsize)
# data_loader = data.DataLoader(dataset=dataset,
# batch_size=batchsize,
# shuffle=shuffle,
# num_workers=num_workers,
# pin_memory=pin_memory)
# return data_loader
class test_dataset:
def __init__(self, image_root, depth_root, testsize):
self.testsize = testsize
self.images = [image_root + f for f in os.listdir(image_root) if f.endswith('.jpg')]
self.depths = [depth_root + f for f in os.listdir(depth_root) if f.endswith('.bmp')
or f.endswith('.png')]
self.images = sorted(self.images)
self.depths = sorted(self.depths)
self.transform = transforms.Compose([
transforms.Resize((self.testsize, self.testsize)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
self.depth_transform = transforms.Compose([
transforms.Resize((self.testsize, self.testsize)),
transforms.ToTensor()])
self.size = len(self.images)
self.index = 0
def load_data(self):
image = self.rgb_loader(self.images[self.index])
depth = self.rgb_loader(self.depths[self.index])
HH = image.size[0]
WW = image.size[1]
image = self.transform(image).unsqueeze(0)
depth = self.depth_transform(depth).unsqueeze(0)
name = self.images[self.index].split('/')[-1]
if name.endswith('.jpg'):
name = name.split('.jpg')[0] + '.png'
self.index += 1
return image, depth, HH, WW, name
def rgb_loader(self, path):
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('RGB')
def binary_loader(self, path):
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('L')