Reproduce “IGRecon4D: Dynamic-Object-Aware Feed-Forward Reconstruction of Driving Scenes” 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

Yifei Tong, Ziyang Huang, Yizhao Han, Xiaoxiao Long, Hao Zhu

IGRecon4D reconstructs driving scenes from recorded sequences using feed-forward Gaussian prediction. Sparse LiDAR guides dense depth that sets Gaussian positions, while instance masks supervise a dynamic-mask head that allocates likely participant regions to per-frame Gaussians; these masks are semantic priors, not motion labels. On Waymo, full-image PSNR is 29.51 versus DGGT’s 26.93; on nuScenes, AnySplat leads full-image metrics.

Summary by alphaXiv.

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