Belval / CRNN

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

max_width 训练的时候是256,预测的时候输入1024行不,这个参数是固定的么 #21

Closed StonyShi closed 5 years ago

StonyShi commented 6 years ago

max_width 训练的时候是256,预测的时候输入1024行不,这个参数是固定的么

zealerww commented 6 years ago

actually, it's ok.but belval set the placeholder to a fixed length inputs = tf.placeholder(tf.float32, [batch_size, max_width, 32, 1]), so image's width must equal to max_width in testing.

you can modify his code, such as inputs = tf.placeholder(tf.float32, [None, None, 32, 1]) and other places should be modified too, you can have a try.

Belval commented 5 years ago

Closing for lack of activity. Feel free to reopen if the above answer was unsatisfactory.