Research map / Point clouds / Point cloud classification
Adaptive Point-Prompt Tuning: Fine-Tuning Heterogeneous Foundation Models for 3D Point Cloud Analysis
Mengke Li, Lihao Chen, Peng Zhang, Yiu‐ming Cheung, Hui Fei Huang
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 2D pre-trained model (ViT-B) to point embedding baseline (ablation) | Point-PN point embedding · ScanObjectNN PB-T50-RS | accuracy: 87.1 -> 90.1 | Improved |
| add 2D pre-trained model (ViT-B) to point embedding baseline (ablation) | Point-PN point embedding · ModelNet40 | accuracy: 93.8 -> 93.9 | No clear effect |
| add position injection (PosIn) (ablation) | PE + 2D · ScanObjectNN PB-T50-RS | accuracy: 90.1 -> 91.2 | Improved |
| add point-prompt tuning (PPT) (ablation) | PE + 2D · ScanObjectNN PB-T50-RS | accuracy: 90.1 -> 91.4 | Improved |
| combine PosIn and PPT (ablation) | PE + 2D · ScanObjectNN PB-T50-RS | accuracy: 91.4 -> 92.6 | Improved |
| combine PosIn and PPT (ablation) | PE + 2D · ModelNet40 | accuracy: 93.9 -> 94.2 | Improved |
| use DeiT instead of ViT-B as 2D prior (ablation) | APPT · ScanObjectNN PB-T50-RS | accuracy: 92.6 -> 88.9 | Hurt |
| use DINOv2 instead of ViT-B as 2D prior (ablation) | APPT · ScanObjectNN PB-T50-RS | accuracy: 92.6 -> 92.6 | No clear effect |
| use RoBERTa text prior instead of CLIP text (ablation) | APPT · ScanObjectNN PB-T50-RS | accuracy: 91.4 -> 87.3 | Hurt |
| use audio prior (ImageBind) instead of ViT-B (ablation) | APPT · ScanObjectNN PB-T50-RS | accuracy: 92.6 -> 88.9 | Hurt |
| replace APF's weight setting with APPT shared-weights prompt generator | APF · ScanObjectNN PB-T50-RS | accuracy: 87.8 -> 92.6 | Improved |
| APPT versus APF in 10-way 20-shot | APF · ModelNet40 10-way 20-shot | accuracy: 95.7 -> 95.3 | Hurt |
Closest papers by meaning
- Point-PEFT: Parameter-Efficient Fine-Tuning for 3D Pre-trained Models (AAAI 2024 · similarity 0.91)
- On Geometry-Enhanced Parameter-Efficient Fine-Tuning for 3D Scene Segmentation (NeurIPS 2025 · similarity 0.89)
- GAPrompt: Geometry-Aware Point Cloud Prompt for 3D Vision Model (ICML 2025 · similarity 0.89)
- Point-PRC: A Prompt Learning Based Regulation Framework for Generalizable Point Cloud Analysis (NeurIPS 2024 · similarity 0.89)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 2025 · similarity 0.88)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.87)
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.
