LWANet

LWANet copied to clipboard

无法复现论文结果

1

同学 您好，我在你的代码上训练了EndoVis2017的数据，对代码的修改有：取消了load_dataset中的padding，将mask//20改为mask//32，以及train中的num_classes。但只能跑到40%+，请问我还需要对代码进行怎样的修改才能达到论文中的效果呢？ 谢谢！

lee1k

Hi, I would like to know which version of pytorch are you using?I would like to reproduce your paper in Endovis 2017, what changes do I need to make?

SWT-1014

Hi, thanks for sharing your excellent work, I'd appreciate if you could introduce how to adjust this model that can use EndoVis 2017. Thanks!

lee1k

The code is not clear

2

How to load the label images???

fcc521zgf