Reproduce “DimFlow: Bidirectional Grid-Expert Learning for Dynamic Gaussian Reconstruction” 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

Gabriel Adineera, Jinoh Kim

DimFlow reconstructs scenes by combining a multiresolution feature grid with a coordinate-based branch that exchange information through adaptive fusion and training-time alignment feedback. Four branches update Gaussian attributes. Across eight D-NeRF scenes, eight HyperNeRF scenes, and ten DAVIS sequences, it reports 39.73 dB PSNR versus 37.81 for 4D-GS at 85 FPS for both, in this evaluation.

Summary by alphaXiv.

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