CloudMamba: Grouped Selective State Spaces for Point Cloud Analysis

2025-11-11

AI Summary

Existing Mamba-inspired point cloud methods suffer from insufficient sequence modeling, weak high-level geometric perception, and overfitting of the Selective State Space (S6) model. To address these issues, we propose CloudMamba—a lightweight and efficient state space network tailored for unordered point clouds. Our key contributions are: (1) a parameter-free multi-axis fusion mechanism with sequence expansion to enhance causal modeling adaptability; (2) a Grouped S6 (GS6) module that mitigates overfitting while preserving selective state modeling capability; and (3) a chained bidirectional scanning architecture that strengthens long-range dependency capture and bidirectional geometric awareness. Evaluated on standard point cloud benchmarks, CloudMamba achieves state-of-the-art performance in classification and segmentation tasks, while significantly reducing parameter count and computational cost—striking an optimal balance between accuracy and efficiency.

Technology Category

• Machine Learning: Learning with Manifolds
• Computer Vision: Multi-modal Vision
• Intelligent Robots: State Estimation

Application Category

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

Abstract

Due to the long-range modeling ability and linear complexity property, Mamba has attracted considerable attention in point cloud analysis. Despite some interesting progress, related work still suffers from imperfect point cloud serialization, insufficient high-level geometric perception, and overfitting of the selective state space model (S6) at the core of Mamba. To this end, we resort to an SSM-based point cloud network termed CloudMamba to address the above challenges. Specifically, we propose sequence expanding and sequence merging, where the former serializes points along each axis separately and the latter serves to fuse the corresponding higher-order features causally inferred from different sequences, enabling unordered point sets to adapt more stably to the causal nature of Mamba without parameters. Meanwhile, we design chainedMamba that chains the forward and backward processes in the parallel bidirectional Mamba, capturing high-level geometric information during scanning. In addition, we propose a grouped selective state space model (GS6) via parameter sharing on S6, alleviating the overfitting problem caused by the computational mode in S6. Experiments on various point cloud tasks validate CloudMamba's ability to achieve state-of-the-art results with significantly less complexity.

Problem

• Addresses imperfect point cloud serialization in Mamba models
• Improves high-level geometric perception through bidirectional scanning
• Mitigates overfitting in selective state space models via parameter sharing

Innovation

• Sequence expanding and merging for unordered point adaptation
• ChainedMamba capturing high-level geometric information
• Grouped selective state space model alleviating overfitting