Reproduce “SplaTraj: Camera Trajectory Generation with Semantic Gaussian Splatting” 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
Xinyi Liu, Tianyi Zhang, Matthew Johnson-Roberson, Weiming Zhi
SplaTraj turns ordered object queries into camera trajectories through language-annotated Gaussian Splatting scenes. It matches queries to scene features, renders target masks, and optimizes a continuous radial basis function (RBF) path through differentiable rendering. RBF achieves the highest intersection-over-union (IoU) in three scenes, including 0.525 ± 0.188 in Kitchen, but polynomial paths have lower jerk in all three; physical robot feasibility is untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.