Reproduce “Learning the 4D World through Point Trajectory Forecasting” 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
Byungwoo Jeon, Sunghwan Hong, Dohyeon Kim, Jonghoon Lee, Jungwoo Park, Inhee Lee, Doosan Baek, Jinil Kim and 5 more
VPM forecasts multi-view video and persistent 3D point trajectories together, using selected points to link motion across frames and camera views. Its trajectory supervision improves both video fidelity and 3D-motion metrics over a same-budget video-only model. On VP-WorldBench’s simulated Robot split, FVD falls from 78.52 to 67.35 and δavg³ᴰ rises from 43.62 to 50.05; VPM assumes calibrated cameras and annotated points.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.