-
Notifications
You must be signed in to change notification settings - Fork 313
/
crnn_model.py
139 lines (115 loc) · 6.86 KB
/
crnn_model.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
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
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time : 17-9-21 下午6:39
# @Author : MaybeShewill-CV
# @Site : https://github.com/MaybeShewill-CV/CRNN_Tensorflow
# @File : crnn_net.py
# @IDE: PyCharm Community Edition
"""
Implement the crnn model mentioned in An End-to-End Trainable Neural Network for Image-based Sequence
Recognition and Its Application to Scene Text Recognition paper
"""
import numpy as np
import tensorflow as tf
from tensorflow.contrib import rnn
from recognize_process.crnn_model import cnn_basenet
from recognize_process.config import model_config
CFG = model_config.cfg
class ShadowNet(cnn_basenet.CNNBaseModel):
def __init__(self, phase, hidden_nums, layers_nums, num_classes):
super(ShadowNet, self).__init__()
if phase == 'train':
self._phase = tf.constant('train', dtype=tf.string)
else:
self._phase = tf.constant('test', dtype=tf.string)
self._hidden_nums = hidden_nums
self._layers_nums = layers_nums
self._num_classes = num_classes
self._is_training = self._init_phase()
def _init_phase(self):
return tf.equal(self._phase, tf.constant('train', dtype=tf.string))
def _conv_stage(self, inputdata, out_dims, name):
with tf.variable_scope(name_or_scope=name):
conv = self.conv2d(inputdata=inputdata, out_channel=out_dims, \
kernel_size=3, stride=1, use_bias=True, name='conv')
bn = self.layerbn(inputdata=conv, is_training=self._is_training, name='bn')
relu = self.relu( inputdata=bn, name='relu')
max_pool = self.maxpooling(inputdata=relu, kernel_size=2, stride=2, name='max_pool')
return max_pool
def _feature_sequence_extraction(self, inputdata, name):
with tf.variable_scope(name_or_scope=name):
conv1 = self._conv_stage(inputdata=inputdata, out_dims=64, name='conv1')
conv2 = self._conv_stage(inputdata=conv1, out_dims=128, name='conv2')
conv3 = self.conv2d(inputdata=conv2, out_channel=256, kernel_size=3, \
stride=1, use_bias=False, name='conv3')
bn3 = self.layerbn(inputdata=conv3, is_training=self._is_training, name='bn3')
relu3 = self.relu(inputdata=bn3, name='relu3')
conv4 = self.conv2d(inputdata=relu3, out_channel=256, kernel_size=3, \
stride=1, use_bias=False, name='conv4')
bn4 = self.layerbn(inputdata=conv4, is_training=self._is_training, name='bn4')
relu4 = self.relu(inputdata=bn4, name='relu4')
max_pool4 = self.maxpooling(inputdata=relu4, kernel_size=[2, 1], \
stride=[2, 1], padding='VALID', name='max_pool4')
conv5 = self.conv2d(inputdata=max_pool4, out_channel=512, kernel_size=3, \
stride=1, use_bias=False, name='conv5')
bn5 = self.layerbn(inputdata=conv5, is_training=self._is_training, name='bn5')
relu5 = self.relu(inputdata=bn5, name='bn5')
conv6 = self.conv2d(inputdata=relu5, out_channel=512, kernel_size=3, \
stride=1, use_bias=False, name='conv6')
bn6 = self.layerbn(inputdata=conv6, is_training=self._is_training, name='bn6')
relu6 = self.relu(inputdata=bn6, name='relu6')
max_pool6 = self.maxpooling(inputdata=relu6, kernel_size=[2, 1], \
stride=[2, 1], name='max_pool6')
conv7 = self.conv2d(inputdata=max_pool6, out_channel=512, kernel_size=2, \
stride=[2, 1], use_bias=False, name='conv7')
bn7 = self.layerbn(inputdata=conv7, is_training=self._is_training, name='bn7')
relu7 = self.relu(inputdata=bn7, name='bn7')
return relu7
def _map_to_sequence(self, inputdata, name):
with tf.variable_scope(name_or_scope=name):
shape = inputdata.get_shape().as_list()
assert shape[1] == 1 # H of the feature map must equal to 1
ret = self.squeeze(inputdata=inputdata, axis=1, name='squeeze')
return ret
def _sequence_label(self, inputdata, name):
with tf.variable_scope(name_or_scope=name):
fw_cell_list = [tf.nn.rnn_cell.LSTMCell(nh, forget_bias=1.0) for
nh in [self._hidden_nums] * self._layers_nums]
# Backward direction cells
bw_cell_list = [tf.nn.rnn_cell.LSTMCell(nh, forget_bias=1.0) for
nh in [self._hidden_nums] * self._layers_nums]
stack_lstm_layer, _, _ = rnn.stack_bidirectional_dynamic_rnn(fw_cell_list,
bw_cell_list, inputdata, #sequence_length=CFG.ARCH.SEQ_LENGTH * np.ones(CFG.TRAIN.BATCH_SIZE),
dtype=tf.float32)
#stack_lstm_layer = self.dropout(inputdata=stack_lstm_layer, keep_prob=0.5,\
# is_training=self._is_training, name='sequence_drop_out')
[batch_s, _, hidden_nums] = inputdata.get_shape().as_list() # [batch, width, 2*n_hidden]
shape = tf.shape(stack_lstm_layer)
rnn_reshaped = tf.reshape(stack_lstm_layer, [shape[0] * shape[1], shape[2]])
w = tf.get_variable(name='w',shape=[hidden_nums, self._num_classes],\
initializer=tf.truncated_normal_initializer(stddev=0.02),trainable=True)
# Doing the affine projection
logits = tf.matmul(rnn_reshaped, w, name='logits')
logits = tf.reshape(logits, [shape[0], shape[1], self._num_classes], name='logits_reshape')
raw_pred = tf.argmax(tf.nn.softmax(logits), axis=2, name='raw_prediction')
# Swap batch and batch axis
rnn_out = tf.transpose(logits, [1, 0, 2], name='transpose_time_major') # [width, batch, n_classes]
return rnn_out, raw_pred
def inference(self, inputdata, name, reuse=False):
with tf.variable_scope(name_or_scope=name, reuse=reuse):
# first apply the cnn feature extraction stage
cnn_out = self._feature_sequence_extraction(
inputdata=inputdata, name='feature_extraction_module')
# second apply the map to sequence stage
sequence = self._map_to_sequence(
inputdata=cnn_out, name='map_to_sequence_module')
# third apply the sequence label stage
net_out, raw_pred = self._sequence_label(
inputdata=sequence, name='sequence_rnn_module')
return net_out
def compute_loss(self, inputdata, labels, name, reuse):
inference_ret = self.inference(
inputdata=inputdata, name=name, reuse=reuse)
loss = tf.reduce_mean(tf.nn.ctc_loss(labels=labels, inputs=inference_ret, sequence_length=\
CFG.ARCH.SEQ_LENGTH * np.ones(CFG.TRAIN.BATCH_SIZE), ignore_longer_outputs_than_inputs=True), name='ctc_loss')
return inference_ret, loss