Presentation

<
Pixal3D: Pixel-Aligned 3D Generation from Images
SessionGenerative 3D & Editing
DescriptionPixal3D is a pixel-aligned paradigm that solves the fidelity bottleneck in image-to-3D synthesis. By utilizing a back-projection conditioning scheme to explicitly lift 2D features into view-aligned 3D volumes, it eliminates ambiguity, enabling scalable generative reconstruction of highly accurate 3D assets and scenes from single or multi-view images.
Authors
Dong-Yang Li
Tsinghua University
Wang Zhao
Tencent
Yuxin Chen
Tencent
Wenbo Hu
Tencent
Meng-Hao Guo
Tsinghua University
Fang-Lue Zhang
Victoria University of Wellington
Ying Shan
Tencent
Shi-Min Hu
Tsinghua University
Event Type
Technical Paper
TimeMonday, 20 July 202611:05am - 11:15am PDT
LocationRoom 408 B
Session Time & Location
Sunday, 19 July 20266:00pm - 8:45pm PDTHall K
Monday, 20 July 202610:45am - 12:35pm PDTRoom 408 B
Keywords
Geometry
Modeling
Registration Categories
Full Conference Supporter
Full Conference
Next PresentationNext PresentationInvSculpt: Inverse Sculpting Modeling via Controlled 3D Generation and a Vector Displacement Field
Similar Presentations
Mix3R: Mixing Feed-forward Reconstruction and Generative 3D Priors for Joint Multi-view Aligned 3D Reconstruction and Pose Estimation
Sparse-to-Complete: From Sparse Image Captures to Complete 3D Scenes
HairGPT: A Unified Autoregressive Framework for 3D Realistic Hairstyle Synthesis