Reproduce “Infer3D: Inverse Generative Modeling Enables Robust 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
Amani Kiruga, Mark Hamilton, Joshua B. Tenenbaum, William T. Freeman, Vincent Sitzmann, Yilun Du, Ayush Tewari
Infer3D reconstructs 3D scenes from single images by searching for a plausible scene and camera that render an explanation of the input. It optimizes latent and rendering parameters while keeping the generator and 2D-to-3D lifter frozen. On ShapeNet Cars with OOD SO(3) viewpoints, it reaches 20.29 dB PSNR, but optimization is costlier and less accurate than feedforward methods on in-distribution data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.