Reproduce “GradSplat: Gradients Inside the Forward Pass for Feed-Forward 3D 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
Jingong Chen, Linlian Jiang, Guangyuan Li, Xinxin Zuo, Yang Wang, Chulwoo Pack, Kyle Gao, Kwanghee Won
GradSplat refines feed-forward 3D Gaussian scenes by rendering them into supplied input views, computing photometric-loss gradients through the rasterizer, and applying three appearance and five shape updates. On the six-view narrow RealEstate10K protocol, it reaches 33.683 dB PSNR versus 31.418 dB for DepthSplat. It leads PSNR and SSIM on both protocols, but YoNoSplat leads wide-protocol LPIPS.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.