Reproduce “No Pose, No Problem in 4D: Feed-Forward Dynamic Gaussians from Unposed Multi-View Videos” 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

Matteo Balice, Yanik Kunzi, Chenyangguang Zhang, Matteo Matteucci, Marc Pollefeys, Sunghwan Hong

NoPo4D reconstructs dynamic 3D scenes from synchronized, uncalibrated multi-camera video in one feed-forward pass. It predicts image-plane pixel shifts and depth changes, using optical-flow pseudo-labels to supervise motion without known camera poses or 3D motion labels. On ExoRecon, it achieves 29.15 PSNR versus 20.44 for matched-training DGGT; it assumes a static camera rig, and quality degrades with moving cameras.

Summary by alphaXiv.

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