Reproduce “Deep Video Matting via Spatio-Temporal Alignment and Aggregation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yanan Sun, Guanzhi Wang, Qiao Gu, Chi-Keung Tang, Yu-Wing Tai

The authors present a video matting system that uses neighboring-frame features to estimate alpha mattes without explicit optical-flow estimation in the matting network. It aligns and selects temporal features and propagates user trimaps from occasional marked frames. On a composited test set, it reports SAD 40.91 with trimaps every frame and 43.66 every 20 frames, while sparser input worsens results.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.