Research map / 3D vision / Face and shape
RI-Mamba: Rotation-Invariant Mamba for Robust Text-to-Shape Retrieval
Khanh Nguyen, Dasith de Silva Edirimuni, Ghulam Mubashar Hassan, Ajmal Mian
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 Hilbert sorting (ablation) | RI-Mamba · Omni3D (ShapeNet pretraining) | accuracy: 13.9 -> 14.7 | Improved |
| add reference frame disambiguation (RFD) (ablation) | RI-Mamba · Omni3D (ShapeNet pretraining) | accuracy: 14.1 -> 14.7 | Improved |
| add positional and orientational embeddings with FiLM (ablation) | RI-Mamba · Omni3D (ShapeNet pretraining) | accuracy: 8.2 -> 14.7 | Improved |
| add adaptive FiLM to pos and ori embeddings (ablation) | RI-Mamba · Omni3D (ShapeNet pretraining) | accuracy: 13.1 -> 14.7 | Improved |
| add bidirectional scanning (ablation) | RI-Mamba · Omni3D (ShapeNet pretraining) | accuracy: 12.7 -> 14.7 | Improved |
| use RI-Mamba instead of SCA3D under random rotation | RI-Mamba · Text2Shape-SO(3) | RR@1: 2.24 -> 13.20 | Improved |
| use RI-Mamba instead of SCA3D (which uses part labels) on canonical poses | RI-Mamba · Text2Shape | RR@5: 35.11 -> 34.57 | No clear effect |
| apply random rotation to database for non-RI DuoMamba | DuoMamba · OmniObject3D text-to-3D (Ensemble) | RR@1: 15.82 -> 7.83 | Hurt |
| scale pretraining data from ShapeNet to Ensemble | RI-Mamba · OmniObject3D text-to-3D | RR@1: 8.39 -> 19.02 | Improved |
| use RI-Mamba instead of RI-Transformer | RI-Mamba · OmniObject3D-SO(3) text-to-3D (Ensemble) | RR@1: 10.39 -> 19.34 | Improved |
| use RI-Mamba instead of RI-Transformer on ModelNet40 classification | RI-Mamba · ModelNet40 (Ensemble) | accuracy: 78.3 -> 76.7 | Hurt |
| use RI-Mamba instead of PointBERT-TAMM on canonical 3D-to-3D retrieval | RI-Mamba · OmniObject3D 3D-to-3D I/I (Ensemble) | mAP: 49.84 -> 47.58 | Hurt |
Closest papers by meaning
- Spectral Informed Mamba for Robust Point Cloud Processing (CVPR 2025 · similarity 0.85)
- 3DET-Mamba: Causal Sequence Modelling for End-to-End 3D Object Detection (NeurIPS 2024 · similarity 0.85)
- Tango3D: Towards Alignment for Global and Local 2D-3D Correspondence (NeurIPS 2026 · similarity 0.85)
- Voxel Mamba: Group-Free State Space Models for Point Cloud based 3D Object Detection (NeurIPS 2024 · similarity 0.85)
- PointMamba: A Simple State Space Model for Point Cloud Analysis (NeurIPS 2024 · similarity 0.85)
- Point Cloud Mamba: Point Cloud Learning via State Space Model (AAAI 2025 · similarity 0.85)
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.
