Reproduce “PanDWM: Panoramic 4D Driving World Models” 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

Jiajun Sun, Zeyu Zhang, Zeyu Ren, Guangfeng Jiang, Yang Zhao, Hao Tang

PanDWM predicts scenes as one 360-degree panoramic band and uses each pixel’s calibrated camera ray to lift predictions into a renderable Gaussian scene. It generates future windows in four student steps. On Val-150, its FVD-1 is 37.4 versus 47.7 for its 50-step teacher, with FID matching the teacher; generation takes 1.96 versus 33.7 seconds per frame in panoramic-band evaluation.

Summary by alphaXiv.

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