PWC-Net+ model weights for Pytorch

[JingleiSHI]

@deqings Dear Sun, Thanks a lot for your excellent work, it really helps me on my current work. I found that you have submitted new pretrained models of your new paper (PWC-Net+), which have better performances. However, there is only caffe version of model weights. Could you please release a pytorch version of pretrained models? Thank you for your attention and looking forward to your response.

Yours sincerely, Jinglei

[AliKafaei]

I am also waiting for that. Hope it would be released soon.

[jrenzhile]

@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]

Is this the link to download PWC NET + weights? https://research.nvidia.com/sites/default/files/pubs/2019-01_A-Fusion-Approach/pwc_net.pth_.tar_.zip Do PWC-NET and PWC-NET + have the same model?

[jrenzhile]

@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]

Thanks @Iuvegood. The results are very similar. For my application I need the network to estimate large displacement. In your code you specify maximum displacement =4 and if we change this valye the pre trained network does not work. Is there any pre trained network for larger maximum displacement for example 5 or 6?

[jrenzhile]

@AliKafaei we don't have the trained network for this.