Yang Yang

设置max_obj=314, k=600能有提升

这是我的调优结果 

 Thanks for your nice work. I wonder how did you distinguish between the left and right hand? As the images below show, the inconsistencies in orientation make it difficult...

When I tried to fix the bugs, I had tried： 1. Visualize Found the demo_vis.png filled with all red pixels. 2.Train only two imgs, and eval them. Found the bugs...

Here is my code: https://github.com/YangYangGirl/SegmenTron/blob/master/segmentron/data/dataloader/mots.py

 https://github.com/LikeLy-Journey/SegmenTron/blob/4bc605eedde7d680314f63d329277b73f83b1c5f/segmentron/data/dataloader/ade.py#L73 Why do you minus 1? Maybe it's the key to solve the problem.

 By getting rid of the minus one operation, the model works !

  Thank you for your reply. For further verification, I downloaded the data set of ade and checked that its mask started from 0. Your explanation may not be...

I found this paper inconsistent with the paper it cited. Have you reproduced his results yet ?