Upload dataset.py
Browse files- dataset.py +60 -0
dataset.py
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import cv2
|
3 |
+
import torch
|
4 |
+
import albumentations as A
|
5 |
+
|
6 |
+
import config as CFG
|
7 |
+
|
8 |
+
|
9 |
+
class CLIPDataset(torch.utils.data.Dataset):
|
10 |
+
def __init__(self, image_filenames, captions, tokenizer, transforms):
|
11 |
+
"""
|
12 |
+
image_filenames and cpations must have the same length; so, if there are
|
13 |
+
multiple captions for each image, the image_filenames must have repetitive
|
14 |
+
file names
|
15 |
+
"""
|
16 |
+
|
17 |
+
self.image_filenames = image_filenames
|
18 |
+
self.captions = list(captions)
|
19 |
+
self.encoded_captions = tokenizer(
|
20 |
+
list(captions), padding=True, truncation=True, max_length=CFG.max_length
|
21 |
+
)
|
22 |
+
self.transforms = transforms
|
23 |
+
|
24 |
+
def __getitem__(self, idx):
|
25 |
+
item = {
|
26 |
+
key: torch.tensor(values[idx])
|
27 |
+
for key, values in self.encoded_captions.items()
|
28 |
+
}
|
29 |
+
|
30 |
+
image = cv2.imread(f"{CFG.image_path}/{self.image_filenames[idx]}")
|
31 |
+
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
|
32 |
+
image = self.transforms(image=image)['image']
|
33 |
+
item['image'] = torch.tensor(image).permute(2, 0, 1).float()
|
34 |
+
item['caption'] = self.captions[idx]
|
35 |
+
|
36 |
+
return item
|
37 |
+
|
38 |
+
|
39 |
+
def __len__(self):
|
40 |
+
return len(self.captions)
|
41 |
+
|
42 |
+
|
43 |
+
|
44 |
+
def get_transforms(mode="train"):
|
45 |
+
if mode == "train":
|
46 |
+
return A.Compose(
|
47 |
+
[
|
48 |
+
A.Resize(CFG.size, CFG.size, always_apply=True),
|
49 |
+
A.Normalize(max_pixel_value=255.0, always_apply=True),
|
50 |
+
]
|
51 |
+
)
|
52 |
+
else:
|
53 |
+
return A.Compose(
|
54 |
+
[
|
55 |
+
A.Resize(CFG.size, CFG.size, always_apply=True),
|
56 |
+
A.Normalize(max_pixel_value=255.0, always_apply=True),
|
57 |
+
]
|
58 |
+
)
|
59 |
+
|
60 |
+
|