-
Notifications
You must be signed in to change notification settings - Fork 1
/
dataset.py
64 lines (41 loc) · 1.69 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
import os
import torch
import torchvision
import numpy as np
from PIL import Image, ImageOps
from torch.utils.data import Dataset, DataLoader
class OCRDataset(Dataset):
def __init__(self, root_dir, transform=None):
self.root_dir = root_dir
files = os.listdir(self.root_dir)
self.items = list(set([fname.split('.')[0] for fname in files]))
self.transform = transform
self.labels, self.char_vec = self._build_labels_and_char_vec()
def _build_labels_and_char_vec(self):
char_vec = set([])
labels = []
for item_name in self.items:
with open(os.path.join(self.root_dir, item_name + ".gt.txt"), 'r') as gt:
label = gt.read()
labels.append(label)
char_vec |= set(label)
char_vec = np.sort(list(char_vec))
char_vec = np.append(char_vec, "_") # BLANK character
return np.array(labels), char_vec
def __len__(self):
return len(self.items)
def get_num_classes(self):
return len(self.char_vec)
def get_encoded_label(self, index):
return np.array([np.searchsorted(self.char_vec[:-1],
list(s)) for s in self.labels[index]])
def get_decoded_label(self, label):
return "".join(self.char_vec[label])
def __getitem__(self, idx):
if torch.is_tensor(idx):
idx = idx.tolist()
item_name = self.items[idx]
image = Image.open(os.path.join(self.root_dir, item_name + ".bin.png"))
if self.transform:
image = self.transform(image)
return image, self.get_encoded_label([idx])[0]