cs.CVNov 18, 2026

Infer3D: Inverse Generative Modeling Enables Robust 3D Reconstruction

Authors: Amani Kiruga, Mark Hamilton, Joshua B. Tenenbaum, William T. Freeman, Vincent Sitzmann, Yilun Du, Ayush Tewari

Organizations: Harvard University · Google DeepMind · Massachusetts Institute of Technology · University of Cambridge

Abstract

Humans can understand the 3D structure of scenes even from highly unfamiliar or extreme viewpoints. Cognitive science suggests this generalization stems from a top-down generative inference strategy. In contrast, state-of-the-art 3D reconstruction neural networks typically rely on amortized, feedforward inference. While fast, this pattern-matching approach degrades when faced with out-of-distribution (OOD) scene conditions unseen during training. Inspired by the generative nature of human vision, we propose Infer3D, an inverse generative modeling framework that recovers the latent variables of a generative model to best match the monocular input. To prevent degenerate flat outputs, we couple the generative image prior with a frozen 3D lifting network that constrains the optimization space to a valid, multi-view-consistent 3D manifold. Across a variety of controlled synthetic and real-world tasks, Infer3D demonstrates superior OOD reconstruction accuracy compared to feedforward networks trained on equal or larger datasets. Furthermore, mirroring how human perception requires more time to process confusing stimuli, we introduce an adaptive inference strategy that expends more test-time compute on ambiguous OOD examples while efficiently routing in-distribution inputs. Project website: https://amanikiruga.github.io/infer3d-site/

Figures & tables

Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Free Geometry: Refining 3D Reconstruction from Longer Versions of Itself

    Apr 15, 2026Yuhang Dai, Xingyi YangFeed-Forward 3D ReconstructionMulti-View 3D Reconstruction

  2. OTT3R: Multi-View 3D Reconstruction and Fast Dataset Generation at 1% Compute

    Sep 28, 2026Brandon Leblanc, Charalambos PoullisMulti-View 3D ReconstructionPseudo-Labeling

  3. PixGR: Pixel-Space Generative Decoding for Fast Feed-Forward 3D Reconstruction

    Oct 22, 2026Ranran Huang, Wenjing Bian, Weixun Luo +2Feed-Forward 3D ReconstructionSparse-View 3D Reconstruction