Summary
Pixal3D is a pixel-aligned 3D generation paradigm proposed by Dong-Yang Li, Wang Zhao, and Yuxin Chen (arXiv:2505.07239) to address the fidelity bottleneck in image-to-3D synthesis. The authors argue that current 3D-native generators suffer from an implicit 2D-3D correspondence problem: they synthesize shapes in canonical space and inject image cues via attention, making pixel-to-3D associations ambiguous. Pixal3D instead generates 3D directly in a pixel-aligned manner consistent with input views, using a pixel back-projection conditioning scheme that lifts multi-scale image features into 3D feature volumes, establishing unambiguous pixel-to-3D correspondence. The approach is scalable, produces high-quality 3D assets with fidelity approaching reconstruction-level performance, and naturally extends to multi-view generation by aggregating back-projected feature volumes across views. The paper also demonstrates benefits for scene synthesis with a modular pipeline producing high-fidelity, object-separated 3D scenes from images. Project page: https://ldyang694.github.io/projects/pixal3d/
Paper Overview
- Field: Computer Vision
- Authors: Dong-Yang Li, Wang Zhao, Yuxin Chen
- Published: 2025-05-09
- arXiv: 2505.07239
- Project page: https://ldyang694.github.io/projects/pixal3d/
Summary
Recent advances in 3D generative models have rapidly improved image-to-3D synthesis quality, enabling higher-resolution geometry and more realistic appearance. Yet fidelity — measuring 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 issue: most 3D-native generators synthesize shape in canonical space and inject image cues via attention, leaving pixel-to-3D associations ambiguous.
Key Contributions
- Pixal3D paradigm: Draws inspiration from 3D reconstruction; instead of generating in a canonical pose, it generates 3D directly in a pixel-aligned way, consistent with the input view.
- Pixel back-projection conditioning: Explicitly lifts multi-scale image features into 3D feature volumes, establishing direct, unambiguous pixel-to-3D correspondence.
- High fidelity at scale: Shows the approach is scalable and produces high-quality assets with fidelity approaching reconstruction-level performance.
- Multi-view generation: Naturally extends to multi-view settings by aggregating back-projected feature volumes across views.
- Scene synthesis: Demonstrates a modular pipeline producing high-fidelity, object-separated 3D scenes from images.
Pixal3D is the first demonstration of large-scale 3D-native pixel-aligned generation, offering a new path toward high-fidelity 3D generation of objects or scenes from single- or multi-view images.
---
*Auto-collected on 2026-05-13*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/177619918