Research map / Point clouds / Point cloud classification
Pyramid Point Cloud Transformer for Large-Scale Place Recognition
Le Hui, Hang Yang, Mingmei Cheng, Jin Xie, Jian Yang
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 |
|---|---|---|---|
| replace PointNet++ with the pyramid point transformer (ablation) | PPT-Net · Oxford | AR@1%: 94.4 -> 98.1 | Improved |
| add local k-NN graph embedding (ablation) | PPT-Net · Oxford | AR@1%: 96.0 -> 98.1 | Improved |
| add transformer (ablation) | PPT-Net · Oxford | AR@1%: 97.3 -> 98.1 | Improved |
| use grouped self-attention with G = 8 instead of G = 1 (ablation) | PPT-Net · Oxford | AR@1% and FLOPs: 97.5 -> 98.1 | Improved |
| increase groups from G = 8 to G = 16 (ablation) | PPT-Net · Oxford | AR@1%: 98.1 -> 97.4 | Hurt |
| use all four resolutions in pyramid VLAD instead of a single scale (ablation) | PPT-Net · Oxford | recall | Improved |
| vary k of the local k-NN graph (ablation) | PPT-Net · Oxford | recall | No clear effect |
| use PPT-Net instead of MinkLoc3D | MinkLoc3D · Oxford (trained on baseline dataset) | AR@1%: 95.9 -> 98.1 | Improved |
| use PPT-Net instead of MinkLoc3D with data augmentation | MinkLoc3D with data augmentation · Oxford AR@1 | AR@1: 93.7 -> 93.5 | Hurt |
| add handcrafted features to LPD-Net | LPD-Net · U.S. (in-house) | AR@1%: 85.7 -> 96.0 | Improved |
| add data augmentation to MinkLoc3D | MinkLoc3D · Oxford | AR@1%: 95.9 -> 97.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PCAN — PCAN: 3D Attention Map Learning Using Contextual Information for Point Cloud Based Retrieval (CVPR 2019) · better on 4 of 4 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
- PointNetVLAD: Deep Point Cloud Based Retrieval for Large-Scale Place Recognition (CVPR 2018 · similarity 0.92)
- LPD-Net: 3D Point Cloud Learning for Large-Scale Place Recognition and Environment Analysis (ICCV 2019 · similarity 0.91)
- SOE-Net: A Self-Attention and Orientation Encoding Network for Point Cloud Based Place Recognition (CVPR 2021 · similarity 0.89)
- PCAN: 3D Attention Map Learning Using Contextual Information for Point Cloud Based Retrieval (CVPR 2019 · similarity 0.87)
- HOTFormerLoc: Hierarchical Octree Transformer for Versatile Lidar Place Recognition Across Ground and Aerial Views (CVPR 2025 · similarity 0.86)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.86)
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.
