Research map / Point clouds / Point cloud classification
StruMamba3D: Exploring Structural Mamba for Self-supervised Point Cloud Representation Learning
Chuxin Wang, Yixin Zha, Wenfei Yang, Tianzhu Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add structural SSM block to standard Mamba baseline (ablation) | Mamba baseline · ScanObjectNN PB-T50-RS | accuracy: 87.23 -> 92.09 | Improved |
| add structural SSM block to standard Mamba baseline (ablation) | Mamba baseline · ModelNet40 | accuracy: 91.86 -> 94.45 | Improved |
| add sequence length-adaptive strategy (ablation) | StruMamba3D · ScanObjectNN PB-T50-RS | accuracy: 92.09 -> 92.75 | Improved |
| add state-wise update (MLP of relative offsets) on top of spatial states (ablation) | Structural SSM · ScanObjectNN PB-T50-RS | accuracy: 91.33 -> 92.75 | Improved |
| add spatial state initialization alone (no state-wise update) (ablation) | Mamba with phi(x) · ScanObjectNN PB-T50-RS | accuracy: 90.94 -> 91.33 | No clear effect |
| generate SSM parameters from spatial offsets only, dropping point-feature term (ablation) | Structural SSM · ScanObjectNN PB-T50-RS | accuracy: 92.75 -> 91.12 | Hurt |
| add adaptive state update mechanism alone (ablation) | Pre-training · ScanObjectNN PB-T50-RS | accuracy: 92.09 -> 92.26 | No clear effect |
| add both ASUM and spatial state consistency loss (lambda=2) (ablation) | Pre-training · ScanObjectNN PB-T50-RS | accuracy: 92.09 -> 92.75 | Improved |
| raise consistency loss weight lambda from 2 to 5 (ablation) | Pre-training · ScanObjectNN PB-T50-RS | accuracy: 92.75 -> 92.57 | Hurt |
| add lightweight convolution for spatial states (ablation) | Structural SSM block · ScanObjectNN PB-T50-RS | accuracy: 91.78 -> 92.22 | Improved |
| add bidirectional scanning (ablation) | Structural SSM block · ModelNet40 | accuracy: 94.81 -> 95.06 | Improved |
| add Hilbert serialization to StruMamba3D (ablation) | StruMamba3D · ScanObjectNN PB-T50-RS | accuracy: 92.75 -> 92.81 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports StruMamba3D as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Rethinking Serialization in Linear 3D Vision: Decoupling Anisotropic Geometry from Isotropic Semantics (ICML 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PointMamba — PointMamba: A Simple State Space Model for Point Cloud Analysis (NeurIPS 2024) · better on 7 of 7 shared results
- PointM2AE — Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-training (NeurIPS 2022) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Point Cloud Mamba: Point Cloud Learning via State Space Model (AAAI 2025 · similarity 0.93)
- PointMamba: A Simple State Space Model for Point Cloud Analysis (NeurIPS 2024 · similarity 0.91)
- Pamba: Enhancing Global Interaction in Point Clouds via State Space Model (AAAI 2025 · similarity 0.90)
- Spectral Informed Mamba for Robust Point Cloud Processing (CVPR 2025 · similarity 0.89)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 2025 · similarity 0.88)
- ZigzagPointMamba: Spatial-Semantic Mamba for Point Cloud Understanding (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
