search

Belval / CRNN

A TensorFlow implementation of https://github.com/bgshih/crnn
MIT License
299 stars 101 forks source link

sequence_length(0) <= 15 #6

Closed aa642531 closed 6 years ago

aa642531 commented 6 years ago

hello,when i tried to train the model ,i meet the follows: InvalidArgumentError (see above for traceback): sequence_length(0) <= 15.it happend at code ' loss = tf.nn.ctc_loss(targets, logits, seq_len) ' when feed datas

aa642531 commented 6 years ago

@Belval

Belval commented 6 years ago

Please see https://github.com/Belval/CRNN/issues/5

aa642531 commented 6 years ago

@Belval Hi,these days i found a way to solve this problem,but i am not sure whether it ' s right. after conv and map2seq,before feed into blstm,i add the code: inputs=tf.reshape(inputs,[config.BATCH_SIZE,256,30]) I resize the all images to the fixed width 256,and it can run successfully.

Belval commented 6 years ago

Does it converge?

Belval commented 6 years ago

@aa642531 There was an error in the CNN part that wasn't returning the right dimension. Please retry with master.