opconty / pytorch_ctpn

This is a pytorch implementation of CTPN(Detecting Text in Natural Image with Connectionist Text Proposal Network). You may want to finetune from: https://drive.google.com/open?id=1JHhI4sEIXfs5gDa1I9AgJBY477HTzAd0

Home Page:https://mp.weixin.qq.com/s/VO42GzwwJBOabpPJOWVn4g

Repository from Github https://github.comopconty/pytorch_ctpnRepository from Github https://github.comopconty/pytorch_ctpn

loss=0.44 on the SROIE dataset

opened this issue · comments

@opconty
I have been training for 1 day on the SROIE dataset., the loss is still 0.44 !
It works well on other datasets, but not the SROIE dataset?
Am I doing something wrong?
dataset download link

X00016469620
X00016469622

Hi,
Could you share your an annotation example?just send me by email: gao.gzhou#gmail.com.

Or you can try my new text detection methods:
std
std++

I think std will work better in this task.

Best

@opconty
sample.zip
or you can download the entire dataset

Why stuck at 0.44 loss?
What are the limitations of pytorch_ctpn?

ok, I'll check that when I have time

to be continued in courao/ocr.pytorch#6