fh2019ustc / DocTr-Plus

The official code for “Deep Unrestricted Document Image Rectification”, TMM, 2023.
https://project.doctrp.top/
MIT License
375 stars 40 forks source link

transformer结构不匹配 #10

Open chenjjcccc opened 8 months ago

chenjjcccc commented 8 months ago

作者您好,我阅读了GeoTr结构框架,发现每一层encoder和decoder都用了两层的MultiheadAttention,论文中说encoder是标准的transformer框架中的encoder,但是transformer中的encoder只用了一层的MultiheadAttention,decoder用了两层。因为我最近在做论文复现的工作,想咨询一下,期待回复!

fh2019ustc commented 8 months ago

以我release的代码为准

BigTree-git commented 7 months ago

可以加个微信交流下心得吗? 我最近也在复现该论文~

lerndeep commented 6 months ago

@BigTree-git @chenjjcccc could you please share your code to reproduce this paper?

I also wants to test on my dataset.