wushilian / STN_CNN_LSTM_CTC_TensorFlow

use STN+CNN+BLSTM+CTC to do OCR
92 stars 27 forks source link

about train set #5

Open SJHBXShub opened 6 years ago

SJHBXShub commented 6 years ago

Hello,I am training this model and I found it hard to converge. Could you tell me how many pictures to train this model and the number of iteration? And what is syn90k, where can I download it? If possible, could you update your final checkpoint. Thank you very much

wushilian commented 6 years ago

@SJHBXShub http://www.robots.ox.ac.uk/~vgg/data/text/

SJHBXShub commented 6 years ago

OK, do you remember how long your train your model and the type of your computer?

wushilian commented 6 years ago

@SJHBXShub 1080Ti and spend 2 or 3days training the model

SJHBXShub commented 6 years ago

Hi, I download syn90K. I found the size of data is more than 9 millions. How many pictures you used to train this model. Have you save the last checkpoint, could you update them. Thank you very much

wushilian commented 6 years ago

@SJHBXShub 800w,and I don't keep the wights of model

SJHBXShub commented 6 years ago

Thank you very much

chl916185 commented 6 years ago

@SJHBXShub ,你的结果是多少呢