Reproduce “ASV3D: Adapting Diffusion-Based Single-View 3D Reconstruction with Extra Imagery” 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

Y Huynh, Duc Thanh Nguyen, Thao Minh Le, Mohamed Abdelrazek

ASV3D uses an additional image to improve single-image 3D reconstruction, selecting which image conditions the diffusion model for each viewpoint without requiring camera poses or calibration. Its gate compares denoising consistency; an optimized version also adapts at test time. On Wonder3D’s 30-object GSO benchmark, optimized ASV3D reduces Chamfer Distance from 0.0218 to 0.0120, including the extra image and updates.

Summary by alphaXiv.

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