Research map / Point clouds / Point cloud classification
PointTPA: Dynamic Network Parameter Adaptation for 3D Scene Understanding
Siyuan Liu, Chaoqun Zheng, Xin Zhou, Tianrui Feng, Dingkang Liang, Xiang Bai
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 DPP (dynamic parameter projector) to static adapters (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 76.8 -> 77.8 | Improved |
| add SNG grouping on top of DPP (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.8 -> 78.4 | Improved |
| number of bases [4,4,2,2,2] instead of [8,8,4,4,4] (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 78.0 -> 78.4 | Improved |
| number of bases [4,4,2,2,2] instead of [1,1,1,1,1] (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.7 -> 78.4 | Improved |
| use 200 groups instead of 400 in the first stage (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 78.0 -> 78.4 | Improved |
| use grouping instead of no grouping (1*) (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.8 -> 78.4 | Improved |
| scaling factor s=1.0 instead of 0.1 (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 78.1 -> 78.4 | Improved |
| DPLayer on down-projection only instead of both (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.8 -> 78.4 | Improved |
| add DPLayer (dynamic weights) to DPP (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.5 -> 78.4 | Improved |
| use last block of each stage instead of dense DPP (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.8 -> 78.4 | Improved |
| mixed insertion at last block of each stage instead of last block only (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.6 -> 78.4 | Improved |
| use space-filling-curve serialization instead of none (ablation) | PointTPA · ScanNet Val (no decoder) | mIoU: 77.8 -> 78.4 | Improved |
Closest papers by meaning
- Point Ladder Tuning: Parameter-Efficient Hierarchical Adaptation for 3D Point Cloud Understanding (ECCV 2026 · similarity 0.89)
- PointDAN: A Multi-Scale 3D Domain Adaption Network for Point Cloud Representation (NeurIPS 2019 · similarity 0.89)
- Point-PEFT: Parameter-Efficient Fine-Tuning for 3D Pre-trained Models (AAAI 2024 · similarity 0.88)
- On Geometry-Enhanced Parameter-Efficient Fine-Tuning for 3D Scene Segmentation (NeurIPS 2025 · similarity 0.87)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.87)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 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.
