Open liyangliu opened 5 years ago
您好 ,pytorch版的 能给个连接吗
@liyangliu can you put your code on github or give access to test your code?
@liyangliu hello, I'm trying to reproduce the sniper under tensorflow, and I meet the same problem with you. Have you found what's the problem? Thanks a lot~
Hi, @bharatsingh430 @mahyarnajibi
I am trying to reproduce sniper under PyTorch, I have checked that my chip is cropped out similar to yours, and my training loss is under proper value. However when I use the trained model to do visualization, there will be a lot of part rois to be classified as objects. Could you please give me some suggestions? Thank you.
BTW, I have successfully trained your mxnet model successfully on the same dataset (VOC2007), your results are visually better in terms of suppression of false positives of parts.