-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
69 lines (58 loc) · 2.79 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
import numpy as np
import torch
from torch.utils.data import Dataset, DataLoader
class HandwritingDataset(Dataset):
def __init__(self, base_path='data/', is_training=True):
self.base_path = base_path
self.char_to_code = torch.load(base_path + 'char_to_code.pt')
self._load_data(is_training)
self._to_one_hot(self.sents, len(self.char_to_code) + 1)
self._summary()
def __getitem__(self, idx):
return self.strks[idx], self.strks_m[idx], \
self.sents[idx], self.sents_m[idx], \
self.onehots[idx]
def __len__(self):
return self.len
def _to_one_hot(self, data, alphabet_len):
onehots = []
for line in data:
oh = np.zeros((line.shape[0], alphabet_len))
oh[np.arange(line.shape[0]), line.int()] = 1
onehots.append(oh)
self.onehots = self._to_torch(np.asarray(onehots))
def _load_data(self, is_t):
if is_t:
self.strks = self._to_torch(np.load(self.base_path + 't_strokes.npy'))
self.sents = self._to_torch(np.load(self.base_path + 't_sentences.npy'))
self.strks_m = self._to_torch(np.load(self.base_path + 't_stroke_mask.npy'))
self.sents_m = self._to_torch(np.load(self.base_path + 't_sentences_mask.npy'))
else:
self.strks = self._to_torch(np.load(self.base_path + 'v_strokes.npy'))
self.sents = self._to_torch(np.load(self.base_path + 'v_sentences.npy'))
self.strks_m = self._to_torch(np.load(self.base_path + 'v_stroke_mask.npy'))
self.sents_m = self._to_torch(np.load(self.base_path + 'v_sentences_mask.npy'))
self.len = self.strks.shape[0]
self.strk_dim = self.strks.shape[2]
self.sent_len = self.sents.shape[1]
def _to_torch(self, np_data, dtype=torch.FloatTensor):
return torch.from_numpy(np_data).type(dtype)
def _summary(self):
n = 56
print('-' * n)
print('| Dataset Info')
print('-' * n)
print('| dataset length: ', self.len)
print('| alphabet length: ', len(self.char_to_code))
print('| strokes shape: ', self.strks.shape)
print('| strokes mask shape: ', self.strks_m.shape)
print('| sentences shape: ', self.sents.shape)
print('| sentences mask shape: ', self.sents_m.shape)
print('| sents one_hot shape: ', self.onehots.shape)
print('-' * n)
# t_dataset = HandwritingDataset(is_training=True)
# loader = DataLoader(t_dataset, batch_size=1, shuffle=False, drop_last=True)
# v_dataset = HandwritingDataset(is_training=False)
# print('length of the training set: ', len(t_dataset))
# print('length of the validation set: ', len(v_dataset))
# t_dataset.to_one_hot(t_dataset.sents[:3], 60)