This is my personal implementation of video interpolation method descried in my graduate thesis. It is a two-step method:flow+synthesis, but with nerual networks. Thanks to Simon Niklaus's fantastic papers about video interpolation and his pytorch implementation about pwc-net.
I train the whole network with vimeo interp. dataset. You can find it here[http://toflow.csail.mit.edu/]. Download the training set and start training. You may need to make some changes in train.py.
python3 train.py
python3 test.py