-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
142 lines (108 loc) · 5.24 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
from PIL import Image
from torch.utils import data
import transforms as trans
from torchvision import transforms
import random
import os
def load_list(dataset_name, data_root):
images = []
labels = []
contours = []
img_root = os.path.join(data_root, dataset_name, "train/images")
label_root = os.path.join(data_root, dataset_name, "train/labels")
contour_root = os.path.join(data_root, dataset_name, "train/contours")
for img in os.listdir(img_root):
images.append(os.path.join(img_root, img))
if os.path.exists(os.path.join(label_root, img[:-4] + ".jpg")):
labels.append(os.path.join(label_root, img[:-4] + ".jpg"))
else:
labels.append(os.path.join(label_root, img[:-4] + ".png"))
if os.path.exists(os.path.join(contour_root, img[:-4] + ".jpg")):
contours.append(os.path.join(contour_root, img[:-4] + ".jpg"))
else:
contours.append(os.path.join(contour_root, img[:-4]+".png"))
return images, labels, contours
def load_test_list(dataset_name, data_root):
images = []
img_root = os.path.join(data_root, dataset_name, "test/images")
for img in os.listdir(img_root):
images.append(os.path.join(img_root, img))
return images
class ImageData(data.Dataset):
def __init__(self, dataset_list, data_root, transform, mode, img_size=224, scale_size=None, t_transform=None,
method=None):
self.dataset_name = dataset_list
if mode == 'train':
self.image_path, self.label_path, self.contour_path = load_list(dataset_list, data_root)
else:
self.image_path = load_test_list(dataset_list, data_root)
self.transform = transform
self.t_transform = t_transform
self.mode = mode
self.img_size = img_size
self.scale_size = scale_size
self.method = method
self.height_size = img_size
self.width_size = img_size * 2
def __getitem__(self, item):
fn = self.image_path[item].split('/')
filename = fn[-1]
image = Image.open(self.image_path[item]).convert('RGB')
image_w, image_h = int(image.size[0]), int(image.size[1])
if self.mode == 'train':
label = Image.open(self.label_path[item]).convert('L')
contour = Image.open(self.contour_path[item]).convert('L')
random_size = self.scale_size
new_img = trans.Scale((random_size * 2, random_size * 1))(image)
new_label = trans.Scale((random_size * 2, random_size * 1), interpolation=Image.NEAREST)(label)
new_contour = trans.Scale((random_size * 2, random_size * 1), interpolation=Image.NEAREST)(contour)
# random crop
w, h = new_img.size
if w != self.img_size * 2 and h != self.img_size:
x1 = random.randint(0, w - self.img_size * 2)
y1 = random.randint(0, h - self.img_size * 1)
new_img = new_img.crop((x1, y1, x1 + self.img_size * 2, y1 + self.img_size * 1))
new_label = new_label.crop((x1, y1, x1 + self.img_size * 2, y1 + self.img_size * 1))
new_contour = new_contour.crop((x1, y1, x1 + self.img_size * 2, y1 + self.img_size * 1))
# random flip
if random.random() < 0.5:
new_img = new_img.transpose(Image.FLIP_LEFT_RIGHT)
new_label = new_label.transpose(Image.FLIP_LEFT_RIGHT)
new_contour = new_contour.transpose(Image.FLIP_LEFT_RIGHT)
# new_img = trans.Scale((self.width_size, self.height_size))(new_img)
# new_label = trans.Scale((self.width_size, self.height_size), interpolation=Image.NEAREST)(new_label)
# new_contour = trans.Scale((self.width_size, self.height_size), interpolation=Image.NEAREST)(new_contour)
new_img = self.transform(new_img)
label_224 = self.t_transform(new_label)
contour_224 = self.t_transform(new_contour)
return new_img, label_224, contour_224,
else:
new_img = self.transform(image)
return new_img, image_w, image_h, self.image_path[item]
def __len__(self):
return len(self.image_path)
def get_loader(dataset_list, data_root, img_size, mode='train', method=None, padding=None):
label_size = img_size
height_size = img_size
width_size = img_size * 2
if mode == 'train':
transform = trans.Compose([
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])
t_transform = trans.Compose([
trans.Scale((width_size, height_size), interpolation=Image.NEAREST),
transforms.ToTensor(),
])
else:
transform = trans.Compose([
trans.Scale((width_size, height_size)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), # 处理的是Tensor
])
scale_size = 256
if mode == 'train':
dataset = ImageData(dataset_list, data_root, transform, mode, img_size, scale_size, t_transform, method=method)
else:
dataset = ImageData(dataset_list, data_root, transform, mode, img_size, method=method)
return dataset