Hi all,
In this guide, I have to add the layer under caffe's directory. I think it requires to re-build the caffe's source after I add my layers. Is it possible to write layers in my own project's directory, and instantiate and register it directly?
My real situation is that I am doing some testing work for a pre-trained model. The model uses a distribution of caffe (they have some custom layers that I do not know the implementation details) that only has binaries available. I need to add a custom loss layer for my own objective function. I am wondering if it is possible.
Thanks a lot!
Regards,
Bo Yang