-
Notifications
You must be signed in to change notification settings - Fork 0
/
data.py
39 lines (26 loc) · 859 Bytes
/
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
import torch
from torch.utils.data import Dataset
import torchvision.transforms as T
from PIL import Image
from glob import glob
class VQData(Dataset):
def __init__(self, root):
super().__init__()
self.root = root
self.imgs = glob(f'{self.root}/*/*.jpg')
print(len(self.imgs))
self.transform = T.Compose([
T.Resize((64, 64)),
T.RandomHorizontalFlip(0.05),
T.ToTensor(),
T.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
def __len__(self):
return len(self.imgs)
def __getitem__(self, idx):
try:
img = Image.open(self.imgs[idx])
img = self.transform(img)
return img
except:
return self.__getitem__(idx - 1)