Error when testing about sfa HOT 5 CLOSED

Could you provide your code and models via baiduNetdisk or other methods?

from sfa.

@liang23333 It has been a long time since the last time I used Caffe. Sorry for no good suggestion for your problem.

1. I guess the first problem arises due to the version of Caffe (I previous used v1.0 for this work.)
2. Where is your MEX-file? If in your Caffe, then refer to the above comment.
3. I did not use multi-GPU in the experiments of this work.
4. My code for this work is the same as you see in this repo. I provide an alternative download URL for the caffemodel file (See Readme.md).

Overall, my suggestion is to ensure you have the same environments as my experiments. At least you should try to use Caffe 1.0 and MATLAB 2016b.

As for the datasets, you can get them here with password b6sv.

New! I have also provided a PyTorch implementation of this method (for simplicity), you can go for it (It should be noted that the results reported in the paper are obtained using Caffe+MATLAB).

Good luck!

from sfa.

Thank you very much for your help!!!

from sfa.

@liang23333 Caffemodel BaiduNetDisk sharing is now available here with password u8sd

from sfa.

@lidq92 I think i have known where the error is, the caffemodel i downloaded is about 80M, but the mdoel is provide is 97M. Thank you for your help.

from sfa.