https://github.com/synckey/tensorflow_lstm_ctc_ocr
代码需要简单修改(tensorflow 0.9)
loss = tf.nn.ctc_loss(inputs=logits,
labels=targets,
sequence_length=seq_len)
#修改
# loss = tf.contrib.ctc.ctc_loss(logits, targets, seq_len)
这种方法能够得到怎样的识别率,尚不清楚
https://github.com/synckey/tensorflow_lstm_ctc_ocr
代码需要简单修改(tensorflow 0.9)
loss = tf.nn.ctc_loss(inputs=logits,
labels=targets,
sequence_length=seq_len)
#修改
# loss = tf.contrib.ctc.ctc_loss(logits, targets, seq_len)
这种方法能够得到怎样的识别率,尚不清楚