Reproduce “Time of the Flight of the Gaussians: Optimizing Depth Indirectly in Dynamic Radiance Fields” 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

Runfeng Li, Mikhail Okunev, Zixuan Guo, Anh Ha Duong, Christian Richardt, Matthew O'Toole, James Tompkin

A method for dynamic scene reconstruction combines time-of-flight imaging with 3D Gaussian splatting through occupancy and reflectivity biases, achieving 100x faster optimization and rendering compared to NeRF-based approaches while maintaining accuracy for complex motions and sparse input data.

Key insights

  • Prioritizing changes in occupancy over reflectivity helps conform Gaussians to true scene structure
  • Low reflectivity initialization prevents incorrect Gaussian positioning in low-reflectivity regions
  • Linear interpolation between timesteps effectively models motion in canonical space

Summary by alphaXiv.

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