Reproduce “PixGR: Pixel-Space Generative Decoding for Fast Feed-Forward 3D 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
Ranran Huang, Wenjing Bian, Weixun Luo, Ye Mao, Krystian Mikolajczyk
PixGR reconstructs posed scenes from sparse images by starting target views from noise and iteratively using them with observed views to predict one shared 3D Gaussian representation. On two-view RE10K extrapolation, it reports 26.03 PSNR, versus 23.98 for ReconSplat, but its FID of 2.75 is worse than MVSplat360’s 1.79. These results apply to posed sparse-view settings, not every ambiguous scene.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.