pumpkin123709 / mseconv

Environment

• GPU: GTX1080Ti
• Ubuntu 16.04.4
• CUDA 10.0
• python 3.6
• torch 1.2.0
• torchvision 0.4.0
• cupy 6.2.0
• scipy 1.3.1
• pillow 6.1.0
• numpy 1.17.0

training data

1. Download Vimeo90k training data from vimeo triplet dataset.

For more informaiton about Vimeo90k, please refer to TOFlow.

Two-frame interpolation

1. To interpolate a frame between arbitrary two frames you have, run interpolate_twoframe.py with following command.

   python interpolate_twoframe.py 

2. Then you will have the interpolated output frame.

Results

• Qualitativa results for large motion
• Qualitativa results for occlusion

Additional Results Video

• orignal video

• Interpolation result of video

• Two comparison videos (.results/test1.mp4 ./results/test2.pm4) Top left: original video; top right: interpolated video (MSEConv); bottom left: AdaCoF; bottom right: AdaCoF+.

Acknowledgements

This code is based on Lee/AdaCoF