Simba: Towards High-Fidelity and Geometrically-Consistent Point Cloud Completion via Transformation Diffusion

2025-11-20

AI Summary

Point cloud completion faces the challenge of simultaneously preserving fine-grained geometric details and maintaining global structural consistency. To address this, we propose a diffusion-based generative completion framework. First, we formulate point-wise transformation as a distribution learning task—eliminating instance-specific memorization to enhance generalization. Second, we incorporate symmetry priors to enforce geometric plausibility and structural coherence. Third, we design a hierarchical Mamba architecture that efficiently captures long-range dependencies while enabling high-fidelity upsampling. Our method achieves state-of-the-art performance on PCN, ShapeNet, and KITTI benchmarks, significantly improving detail fidelity, structural completeness, and robustness to input noise.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Point cloud completion is a fundamental task in 3D vision. A persistent challenge in this field is simultaneously preserving fine-grained details present in the input while ensuring the global structural integrity of the completed shape. While recent works leveraging local symmetry transformations via direct regression have significantly improved the preservation of geometric structure details, these methods suffer from two major limitations: (1) These regression-based methods are prone to overfitting which tend to memorize instant-specific transformations instead of learning a generalizable geometric prior. (2) Their reliance on point-wise transformation regression lead to high sensitivity to input noise, severely degrading their robustness and generalization. To address these challenges, we introduce Simba, a novel framework that reformulates point-wise transformation regression as a distribution learning problem. Our approach integrates symmetry priors with the powerful generative capabilities of diffusion models, avoiding instance-specific memorization while capturing robust geometric structures. Additionally, we introduce a hierarchical Mamba-based architecture to achieve high-fidelity upsampling. Extensive experiments across the PCN, ShapeNet, and KITTI benchmarks validate our method's state-of-the-art (SOTA) performance.

Problem

• Preserving fine details while ensuring global structural integrity
• Overcoming overfitting in geometric transformation regression methods
• Enhancing robustness against input noise in point cloud completion

Innovation

• Reformulates transformation regression as distribution learning
• Integrates symmetry priors with diffusion model generation
• Uses hierarchical Mamba architecture for upsampling