The resnet101 performance

Open LaoYang1994 opened this issue 6 years ago • 3 comments

We try to get the performance using resnet101 in your paper. But the final performance is a little worse than the reported result.

I wonder is there any problems in our settings? Do you make sure the code can get the pq about 46？

May 14 '19 08:05 LaoYang1994

ResNet101 results use deformconv for res3-5

May 14 '19 19:05 YuwenXiong

We try to get the performance using resnet101 in your paper. But the final performance is a little worse than the reported result.

I wonder is there any problems in our settings? Do you make sure the code can get the pq about 46？

Hi, Laoyang, could you please help me with my issue? I don;t know does it will influence the data we use? https://github.com/uber-research/UPSNet/issues/53#issue-450141801

May 30 '19 05:05 JoyHuYY1412

@LaoYang1994 Hi, Laoyang, did you retry to train resnet101 with dcn? Can it achieve the reported result?

Aug 14 '19 02:08 dingguo1996