This is my personal implementation of the paper "Phase-based frame interpolation for Video". To be precise, this implementation almost matches this MATLAB implementation. As claimed by the orginal author of the MATLAB code, the implementation is highly un-optimized. Nevertheless, I added GPU support to reduce the processing time while still maintaining the clarity of the method.
Also, if you use this implementation, you must cite the following paper:
S. Meyer, O. Wang, H. Zimmer, M. Grosse and A. Sorkine-Hornung
Phase-based frame interpolation for video
2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
This implementation uses a version of pyrtools by Eero Simoncelli in Python modified by me. Please check this website for more information.
cupy v3.0.0a1 (fft is not supported in the earlier versions)
python demo.py image1_path image2_path --dev cpu --n_frames 1 --save 1 --show 1
Use help to see more options.
