search

liuhu-bigeye / enctc.crnn

Project for Connectionist Temporal Classification with Maximum Entropy Regularization.
MIT License
143 stars 40 forks source link

Request for best performance model #7

Open chocolate9624 opened 4 years ago

chocolate9624 commented 4 years ago

Hi, I really appreciate your wonderful work. But when I run the "segctcent_800w.sh" script, I get much lower Seq_Acc than they are in the paper.

I guess the performance gaps come from the training iteration steps. The scripts (segctcent_800w.sh) says 50 epochs. But in your paper, you says 150 epochs. Am I right ?

It takes almost two months to finish the 50 epochs training steps. It seems impossible for me to finish the left 100 epochs. So could you please provide the best performance model on Synth90K dataset?

I just want to make a comparison and validate my work. Many thanks.

Gaming722 commented 2 months ago

Hello, I want to know where I can download the dataset file "max90k_train_val.lmdb", "data/max90k_test.lmdb" to run the "segctcent_800w.sh" script.