Question about the model
Sam1224 opened this issue · 2 comments
Hi, thanks for your great work!
I have a question related to model/DMNet.py, is current version the correct one aligned with the paper?
Because in L177 of this file, "alpha_conv" is included for optimization, but is not included in the model. Although I can easily fix these errors, I cannot reproduce the reported results in Table 2.
With resnet-50 as the backbone, under 1-shot setting, the reproduced results of fold-0/1/2 of iSAID dataset are 0.5194, 0.4012 and 0.5134, respectively, which are lower than reported. Therefore, I want to check if the current version correct or not.
请问环境要求是什么
@Sam1224 I also find the "alpha_conv" error , and the DMNet performance on the iSAID dataset still low and i also do not reproduce the reported results in the artical.