zimenglan

hi @samuel1208 yes，or you can modify your `concat layer` so that it can deal with different size of feature map.

hi @yeezhu , do you still have a plan to release the caffe version? thanks.

@yeezhu , Thanks for your reply. BTW, in training stage, the SPN does not need bp? do i understand it right?

u should read about those papers for more details. because i just use them and not go deep insight into it.

hi @filmo, do you set the path of caffe correctly in config file?

hi @xuanyuyt, in paper, it's binary. here i just use `softmax`, and all classes share the same `alpha`.

hi @mathmanu, i don't have a plan to add a licence file. so you can use it by yourself for any purpose.

Hi @bailvwangzi, I only test focal-loss on my own dataset, found that it drops a little performance. so sad.

hi @bailvwangzi, `alpha = 1` and `gamma=3` can boost 1.4% mAP on pascal voc 2007 testset using `pvanet`.

hi @bailvwangzi, it's my friend's results. yes, use two stages (PVANET), in which we replace the `softmax loss` with `focal loss` in `Fast RCNN` instead of `RPN`.