Research map / Point clouds / Point cloud classification
SVT-Net: Super Light-Weight Sparse Voxel Transformer for Large Scale Place Recognition
Zhaoxin Fan, Zhenbo Song, Hongyan Liu, Zhiwu Lu, Jun He, Xiaoyong Du
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 both ASVT and CSVT sparse voxel transformers | SVT-Net · U.S. | average recall at top-1 (%): 86.7 -> 90.1 | Improved |
| add both ASVT and CSVT sparse voxel transformers | SVT-Net · R.A. | average recall at top-1 (%): 80.4 -> 84.3 | Improved |
| add both ASVT and CSVT sparse voxel transformers | SVT-Net · B.D. | average recall at top-1 (%): 81.5 -> 85.5 | Improved |
| add CSVT to ASVT-Net (ablation) | SVT-Net · Oxford | average recall at top-1% (%): 98.0 -> 97.8 | No clear effect |
| add CSVT to ASVT-Net (ablation) | SVT-Net · B.D. | average recall at top-1% (%): 88.4 -> 90.7 | Improved |
| add ASVT to CSVT-Net (ablation) | SVT-Net · B.D. | average recall at top-1% (%): 89.5 -> 90.7 | Improved |
| use 4 tokens instead of 8 (ablation) | SVT-Net · Oxford | average recall at top-1% (%): 97.8 -> 97.9 | No clear effect |
| use 6 tokens instead of 8 (ablation) | SVT-Net · Oxford | average recall at top-1% (%): 97.8 -> 98.0 | No clear effect |
| use 10 tokens instead of 8 (ablation) | SVT-Net · U.S. | average recall at top-1 (%): 90.1 -> 87.2 | Hurt |
| reduce descriptor dimension to 128 (ablation) | SVT-Net · R.A. | average recall at top-1 (%): 84.3 -> 81.9 | Hurt |
| increase descriptor dimension to 384 (ablation) | SVT-Net · R.A. | average recall at top-1 (%): 84.3 -> 84.9 | Improved |
| increase descriptor dimension to 512 (ablation) | SVT-Net · B.D. | average recall at top-1% (%): 90.7 -> 88.2 | Hurt |
Later papers that compared with this method
1 later accepted paper reports SVT-Net as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- CASSPR: Cross Attention Single Scan Place Recognition (ICCV 2023 · better on 2 of 2 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
- MsSVT++: Mixed-Scale Sparse Voxel Transformer With Center Voting for 3D Object Detection (TPAMI 2023 · similarity 0.89)
- DSVT: Dynamic Sparse Voxel Transformer With Rotated Sets (CVPR 2023 · similarity 0.88)
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.88)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.85)
- SuperVLAD: Compact and Robust Image Descriptors for Visual Place Recognition (NeurIPS 2024 · similarity 0.85)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.84)
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.
