dotnet
Is it possible to implement Faster rcnn?
I try to use Faster rcnn for object detection, training and inference. Is it possible to implement faster rcnn? How much work is required?
It would take me a few days to implement based on the torchvision code, which isn't too bad.
Testing would be the real challenge -- I don't think that I have the data to test it properly. This is the kind of thing that would be perfect for someone who is knowledgeable about the domain to help out with.
@xhuan8 -- I see that you have been working on things in your private fork. Is there anything ready to PR to dotnet/TorchSharp?
I haven't finished the testing, the torchvision dll seems not working, maybe need implement custom roi_align method. also the packing has some problem on my local machine.
I'm a kind of busy recently, will continue the work on the middle of next month.