Reproduce “CC3D: Layout-Conditioned Generation of Compositional 3D Scenes” 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
Sherwin Bahmani, Jeong Joon Park, Despoina Paschalidou, Xingguang Yan, Gordon Wetzstein, Leonidas Guibas, Andrea Tagliasacchi
CC3D generates multi-object 3D scenes from top-down semantic layouts, using a 2D network to build a 3D feature field that is rendered from camera views. It is trained on single-view RGB images and layouts that need not be paired one-to-one. On 3D-FRONT bedrooms at 256 × 256, it achieves FID 28.5, versus 49.0 for EG3D; layout guidance does not ensure exact adherence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.