File size: 10,555 Bytes
13580fb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
import torch.nn as nn
from util.util import to_device
from torch.nn import init
import os
import torch
from .networks import *
from params import *

class BidirectionalLSTM(nn.Module):

    def __init__(self, nIn, nHidden, nOut):
        super(BidirectionalLSTM, self).__init__()

        self.rnn = nn.LSTM(nIn, nHidden, bidirectional=True)
        self.embedding = nn.Linear(nHidden * 2, nOut)


    def forward(self, input):
        recurrent, _ = self.rnn(input)
        T, b, h = recurrent.size()
        t_rec = recurrent.view(T * b, h)

        output = self.embedding(t_rec)  # [T * b, nOut]
        output = output.view(T, b, -1)

        return output


class CRNN(nn.Module):

    def __init__(self, leakyRelu=False):
        super(CRNN, self).__init__()
        self.name = 'OCR'
        #assert opt.imgH % 16 == 0, 'imgH has to be a multiple of 16'

        ks = [3, 3, 3, 3, 3, 3, 2]
        ps = [1, 1, 1, 1, 1, 1, 0]
        ss = [1, 1, 1, 1, 1, 1, 1]
        nm = [64, 128, 256, 256, 512, 512, 512]

        cnn = nn.Sequential()
        nh = 256
        dealwith_lossnone=False # whether to replace all nan/inf in gradients to zero

        def convRelu(i, batchNormalization=False):
            nIn = 1 if i == 0 else nm[i - 1]
            nOut = nm[i]
            cnn.add_module('conv{0}'.format(i),
                           nn.Conv2d(nIn, nOut, ks[i], ss[i], ps[i]))
            if batchNormalization:
                cnn.add_module('batchnorm{0}'.format(i), nn.BatchNorm2d(nOut))
            if leakyRelu:
                cnn.add_module('relu{0}'.format(i),
                               nn.LeakyReLU(0.2, inplace=True))
            else:
                cnn.add_module('relu{0}'.format(i), nn.ReLU(True))

        convRelu(0)
        cnn.add_module('pooling{0}'.format(0), nn.MaxPool2d(2, 2))  # 64x16x64
        convRelu(1)
        cnn.add_module('pooling{0}'.format(1), nn.MaxPool2d(2, 2))  # 128x8x32
        convRelu(2, True)
        convRelu(3)
        cnn.add_module('pooling{0}'.format(2),
                       nn.MaxPool2d((2, 2), (2, 1), (0, 1)))  # 256x4x16
        convRelu(4, True)
        if resolution==63:
            cnn.add_module('pooling{0}'.format(3),
                           nn.MaxPool2d((2, 2), (2, 1), (0, 1)))  # 256x4x16
        convRelu(5)
        cnn.add_module('pooling{0}'.format(4),
                       nn.MaxPool2d((2, 2), (2, 1), (0, 1)))  # 512x2x16
        convRelu(6, True)  # 512x1x16

        self.cnn = cnn
        self.use_rnn = False
        if self.use_rnn:
            self.rnn = nn.Sequential(
                BidirectionalLSTM(512, nh, nh),
                BidirectionalLSTM(nh, nh, ))
        else:
            self.linear = nn.Linear(512, VOCAB_SIZE)

        # replace all nan/inf in gradients to zero
        if dealwith_lossnone:
            self.register_backward_hook(self.backward_hook)

        self.device = torch.device('cuda:{}'.format(0)) 
        self.init = 'N02'
        # Initialize weights
        
        self = init_weights(self, self.init)

    def forward(self, input):
        # conv features
        conv = self.cnn(input)
        b, c, h, w = conv.size()
        if h!=1:
            print('a')
        assert h == 1, "the height of conv must be 1"
        conv = conv.squeeze(2)
        conv = conv.permute(2, 0, 1)  # [w, b, c]

        if self.use_rnn:
            # rnn features
            output = self.rnn(conv)
        else:
            output = self.linear(conv)
        return output

    def backward_hook(self, module, grad_input, grad_output):
        for g in grad_input:
            g[g != g] = 0  # replace all nan/inf in gradients to zero


class OCRLabelConverter(object):
    """Convert between str and label.

    NOTE:
        Insert `blank` to the alphabet for CTC.

    Args:
        alphabet (str): set of the possible characters.
        ignore_case (bool, default=True): whether or not to ignore all of the case.
    """

    def __init__(self, alphabet, ignore_case=False):
        self._ignore_case = ignore_case
        if self._ignore_case:
            alphabet = alphabet.lower()
        self.alphabet = alphabet + '-'  # for `-1` index

        self.dict = {}
        for i, char in enumerate(alphabet):
            # NOTE: 0 is reserved for 'blank' required by wrap_ctc
            self.dict[char] = i + 1

    def encode(self, text):
        """Support batch or single str.

        Args:
            text (str or list of str): texts to convert.

        Returns:
            torch.IntTensor [length_0 + length_1 + ... length_{n - 1}]: encoded texts.
            torch.IntTensor [n]: length of each text.
        """
        '''
        if isinstance(text, str):
            text = [
                self.dict[char.lower() if self._ignore_case else char]
                for char in text
            ]
            length = [len(text)]
        elif isinstance(text, collections.Iterable):
            length = [len(s) for s in text]
            text = ''.join(text)
            text, _ = self.encode(text)
        return (torch.IntTensor(text), torch.IntTensor(length))
        '''
        length = []
        result = []
        for item in text:
            item = item.decode('utf-8', 'strict')
            length.append(len(item))
            for char in item:
                index = self.dict[char]
                result.append(index)

        text = result
        return (torch.IntTensor(text), torch.IntTensor(length))

    def decode(self, t, length, raw=False):
        """Decode encoded texts back into strs.

        Args:
            torch.IntTensor [length_0 + length_1 + ... length_{n - 1}]: encoded texts.
            torch.IntTensor [n]: length of each text.

        Raises:
            AssertionError: when the texts and its length does not match.

        Returns:
            text (str or list of str): texts to convert.
        """
        if length.numel() == 1:
            length = length[0]
            assert t.numel() == length, "text with length: {} does not match declared length: {}".format(t.numel(),
                                                                                                         length)
            if raw:
                return ''.join([self.alphabet[i - 1] for i in t])
            else:
                char_list = []
                for i in range(length):
                    if t[i] != 0 and (not (i > 0 and t[i - 1] == t[i])):
                        char_list.append(self.alphabet[t[i] - 1])
                return ''.join(char_list)
        else:
            # batch mode
            assert t.numel() == length.sum(), "texts with length: {} does not match declared length: {}".format(
                t.numel(), length.sum())
            texts = []
            index = 0
            for i in range(length.numel()):
                l = length[i]
                texts.append(
                    self.decode(
                        t[index:index + l], torch.IntTensor([l]), raw=raw))
                index += l
            return texts


class strLabelConverter(object):
    """Convert between str and label.
    NOTE:
        Insert `blank` to the alphabet for CTC.
    Args:
        alphabet (str): set of the possible characters.
        ignore_case (bool, default=True): whether or not to ignore all of the case.
    """

    def __init__(self, alphabet, ignore_case=False):
        self._ignore_case = ignore_case
        if self._ignore_case:
            alphabet = alphabet.lower()
        self.alphabet = alphabet + '-'  # for `-1` index

        self.dict = {}
        for i, char in enumerate(alphabet):
            # NOTE: 0 is reserved for 'blank' required by wrap_ctc
            self.dict[char] = i + 1

    def encode(self, text):
        """Support batch or single str.
        Args:
            text (str or list of str): texts to convert.
        Returns:
            torch.IntTensor [length_0 + length_1 + ... length_{n - 1}]: encoded texts.
            torch.IntTensor [n]: length of each text.
        """
        '''
        if isinstance(text, str):
            text = [
                self.dict[char.lower() if self._ignore_case else char]
                for char in text
            ]
            length = [len(text)]
        elif isinstance(text, collections.Iterable):
            length = [len(s) for s in text]
            text = ''.join(text)
            text, _ = self.encode(text)
        return (torch.IntTensor(text), torch.IntTensor(length))
        '''
        length = []
        result = []
        results = []
        for item in text:
            item = item.decode('utf-8', 'strict')
            length.append(len(item))
            for char in item:
                index = self.dict[char]
                result.append(index)
            results.append(result)
            result = []

        return (torch.nn.utils.rnn.pad_sequence([torch.LongTensor(text) for text in results], batch_first=True), torch.IntTensor(length))

    def decode(self, t, length, raw=False):
        """Decode encoded texts back into strs.
        Args:
            torch.IntTensor [length_0 + length_1 + ... length_{n - 1}]: encoded texts.
            torch.IntTensor [n]: length of each text.
        Raises:
            AssertionError: when the texts and its length does not match.
        Returns:
            text (str or list of str): texts to convert.
        """
        if length.numel() == 1:
            length = length[0]
            assert t.numel() == length, "text with length: {} does not match declared length: {}".format(t.numel(),
                                                                                                         length)
            if raw:
                return ''.join([self.alphabet[i - 1] for i in t])
            else:
                char_list = []
                for i in range(length):
                    if t[i] != 0 and (not (i > 0 and t[i - 1] == t[i])):
                        char_list.append(self.alphabet[t[i] - 1])
                return ''.join(char_list)
        else:
            # batch mode
            assert t.numel() == length.sum(), "texts with length: {} does not match declared length: {}".format(
                t.numel(), length.sum())
            texts = []
            index = 0
            for i in range(length.numel()):
                l = length[i]
                texts.append(
                    self.decode(
                        t[index:index + l], torch.IntTensor([l]), raw=raw))
                index += l
            return texts