Review — SepConv++: Revisiting Adaptive Convolutions for Video Frame Interpolation

SepConv++: A Bunch of Small Improvements for Adaptive Separable Convolutions, Achieve SOTA Performance

Kernel-Based Interpolation with Spatially-Varying Kernels


1. Proposed Video Frame Interpolation Framework

Proposed Video Frame Interpolation Framework (φ denotes the adaptive separable convolution operator)

2. Delayed Padding

3. Input Normalization

4. Network Improvements

5. Kernel Normalization

6. Contextual Training

7. Self-Ensembling

8. Experimental Results

Ablation experiments to quantitatively analyze the effects of our proposed techniques
Effect of combining the mean of eight independent predictions for several video frame interpolation methods
Qualitative Comparison of SepConv++ Over SepConv

PhD, Researcher. I share what I've learnt and done. :) My LinkedIn:, My Paper Reading List: