Zhile Ren
Zhile Ren
@kellyJin0505 It says "nvcc: command not found", meaning cuda is not installed in your machine
@JingleiSHI @AliKafaei In a work building on PWC-Net, we used the updated PWC-NET+ model: https://github.com/NVlabs/PWC-Net/tree/master/Multi_Frame_Flow#test. You may try that and see if it can help to improve your prediction.
@AliKafaei they have subtle differences, you may use the code in https://github.com/NVlabs/PWC-Net/tree/master/Multi_Frame_Flow for test.
@AliKafaei we don't have the trained network for this.
@bbidong no we didn't provide the training code. You may refer to the training protocol in caffe.
@tobymu can you explain what you mean "get the mask with flow"?
@tobymu it's two totally different things. The mask in Fei-Fei's paper means the occlusion map, and the "mask" used in this code is more of an implementational hack in PyTorch0.2...
@bkvie My thinking is that you can certainly get creative and design a novel loss function, but could you please elaborate "deform image 1" and "negative correlation"?
We haven't tested on TitanV or higher version of CUDA. But the inference time takes less than a second, excluding the model loading time. Can you check the speed for...
We didn't provide the training code in pytorch, but you can always try using your own data