Overview
Field: Computer Vision Authors: Dong-Yang Li, Wang Zhao, Yuxin Chen Published: 2025-05-09 arXiv: 2505.07239
Abstract (translated)
Recent advances in 3D generative models have rapidly improved image-to-3D synthesis quality, enabling higher-resolution geometry and more realistic appearance. However, fidelity — the pixel-level faithfulness of the generated 3D asset to the input image — remains a central bottleneck. The authors argue this stems from an implicit 2D-3D correspondence problem: most 3D-native generators synthesize shape in canonical space and inject image cues via attention, leaving pixel-to-3D associations ambiguous.
Drawing inspiration from 3D reconstruction, the paper proposes Pixal3D, a pixel-aligned 3D generation paradigm for high-fidelity 3D asset creation from images. Instead of generating in a canonical pose, Pixal3D generates 3D directly in a pixel-aligned way, consistent with the input view.
Key Contributions
- Pixel back-projection conditioning: multi-scale image features are explicitly lifted into 3D feature volumes, establishing direct and unambiguous pixel-to-3D correspondence.
- Scalable, high-fidelity generation: Pixal3D produces high-quality 3D assets with fidelity approaching reconstruction-level accuracy.
- Multi-view generation: naturally extended by aggregating back-projected feature volumes across views.
- Scene synthesis: a modular pipeline produces high-fidelity, object-separated 3D scenes from images.
Project page: https://ldyang694.github.io/projects/pixal3d/
--- *Auto-collected on 2026-05-13*