-
Notifications
You must be signed in to change notification settings - Fork 29
/
dataset.py
executable file
·42 lines (32 loc) · 1.27 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
from io import BytesIO
import lmdb
import numpy as np
from PIL import Image
from PIL import Image
from torch.utils.data import Dataset
class MultiResolutionDataset(Dataset):
def __init__(self, path, transform, resolution=256):
self.env = lmdb.open(path, max_readers=32, readonly=True, lock=False, readahead=False, meminit=False,)
if not self.env:
raise IOError("Cannot open lmdb dataset", path)
with self.env.begin(write=False) as txn:
self.length = int(txn.get("length".encode("utf-8")).decode("utf-8"))
self.resolution = resolution
self.transform = transform
def __len__(self):
return self.length
def __getitem__(self, index):
while True:
try:
with self.env.begin(write=False) as txn:
key = f"{self.resolution}-{str(index).zfill(5)}".encode("utf-8")
img_bytes = txn.get(key)
buffer = BytesIO(img_bytes)
img = Image.open(buffer)
break
except:
print(f"ERROR loading image {index}")
index = int(np.random.rand() * self.length)
print(f"Trying again with {index}...")
img = self.transform(img)
return img