Error about Flownet2 installation

[cherryhanya]

When I tried to execute example command, I encountered with this issues. Could you please help to solve this problem? I guess there is a problem regarding the 'resample2d_cuda.cpython-36m-x86_64-linux-gnu.so' file

[lukas-blecher]

I can't help you with the flownet2 modules. I've also had problems with them. So I removed them and implemented a similar network in its place: The LiteFlowNet (pytorch, original). This implementation doesn't need the flownet2 modules at all but you need to install CuPy

You can check it out here: https://github.com/lukas-blecher/Deep-Flow-Guided-Video-Inpainting

[cherryhanya]

I can't help you with the flownet2 modules. I've also had problems with them. So I removed them and implemented a similar network in its place: The LiteFlowNet (pytorch, original). This implementation doesn't need the flownet2 modules at all but you need to install CuPy

You can check it out here: https://github.com/lukas-blecher/Deep-Flow-Guided-Video-Inpainting

Thank you so much for your comment! It was really helpful for me. I succeed to extract flow from LiteFlowNet. However, when i tried to get the inpainting result again, I got this error from 'FlowInitial.py' part. I tried to change the type of 'tmp_mask_resized' to cv::UMat, but it didn't work. Do you have any idea why it happens?

[lukas-blecher]

seems like there was missing a keyword argument. Try again with interpolation=cv2.INTER_NEAREST or pull the changes.

[hanuilee]

seems like there was missing a keyword argument. Try again with interpolation=cv2.INTER_NEAREST or pull the changes.

Problem solved! Thank you so much for the quick and kind answers!

[yuyang16101066]

Where did you get pretrained_model_liteflownet?

[lukas-blecher]

@yuyang16101066 I've linked them in the README of the fork or in this issue

[yuyang16101066]

@yuyang16101066 I've linked them in the README of the fork or in this issue

Thanks for your reply. I have generated flow with LiteFlowNet.