Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates
Published in The Thirty-Ninth Annual Conference on Neural Information Processing Systems(NeurIPS), Spotlight, 2025
FCM is a training-free, likelihood-guided diffusion update that auto-tunes step sizes via forward-mode autodiff and curvature probes, enabling fast, stable 3D point-cloud reconstruction from single/multi-view inputs—no retraining.
Recommended citation: Shin, Seunghyeok; Kim, Dabin; Lim, Hongki. (2025). Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates. Proceedings of the Thirty-Ninth Conference on Neural Information Processing Systems (NeurIPS 2025).
Download Paper