DualPrim: Compact 3D Reconstruction with Positive and Negative Primitives

IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR) 2026

Xiaoxu Meng^1,* Independent Researcher
Zhongmin Chen^4,5,* ICT, CAS and UCAS
Bo Yang² Waymo LLC
Weikai Chen^3,† Independent Researcher
Weixiao Liu⁴ Lucid Motors
Lin Gao⁴ ICT, CAS

^* Equal contribution • ^† Corresponding author

DualPrim is a differentiable representation that models 3D shapes using dual-primitives composed of positive- and negative-density superquadrics. This additive-subtractive design increases the representational power (e.g. holes and concavities) without sacrificing compactness or differentiability. Integrated into a volumetric differentiable renderer, DualPrim supports end-to-end learning from multi-view images and enables seamless mesh extraction through closed-form Boolean differencing.

Abstract

Neural reconstructions often trade structure for fidelity, yielding dense and unstructured meshes with irregular topology and weak part boundaries that hinder editing, animation, and downstream asset reuse. We present DualPrim, a compact and structured 3D reconstruction framework. Unlike additive-only implicit or primitive methods, DualPrim represents shapes with positive and negative superquadrics: the former builds the bases while the latter carves local volumes through a differentiable operator, enabling topology-aware modeling of holes and concavities. This additive-subtractive design increases the representational power without sacrificing compactness or differentiability.

Presentation Video

YouTube Source

Bilibili Source

Resources

📄 Paper (Coming Soon) 💻 Code (Coming Soon)

BibTeX

Coming Soon