Reproduce “OccJEPA: Occlusion as Masking for Amodal Video Segmentation” 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

Genpei Zhang, Runmin Jiang, Xingzu Zhan, Yibo Zhong, Ruien Li, Zhou Zhang

OccJEPA completes objects’ outlines across video, including regions hidden by occlusion. It treats tokens outside the visible mask as missing video information, uses a pretrained predictor to infer their features, and decodes those features into masks. On a held-out TAO-Amodal test split, it scores 92.99 AP50 versus 88.99 for Diffusion-VAS, measuring box—not mask—accuracy.

Summary by alphaXiv.

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