Research map / Point clouds / Point cloud classification
Parameter-Efficient Fine-Tuning in Spectral Domain for Point Cloud Learning
Dingkang Liang, Tianrui Feng, Xin Zhou, Yumeng Zhang, Zhikang Zou, 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 |
|---|---|---|---|
| replace full fine-tuning with PointGST spectral tuning | PointGPT-L · ScanObjectNN PB_T50_RS | overall accuracy: 93.4 -> 94.83 | Improved |
| replace full fine-tuning with PointGST | Point-BERT · ScanObjectNN OBJ_BG | overall accuracy: 87.43 -> 91.39 | Improved |
| replace full fine-tuning with PointGST | Point-MAE · ScanObjectNN PB_T50_RS | overall accuracy: 85.18 -> 85.29 | No clear effect |
| replace full fine-tuning with PointGST | ReCon · ScanObjectNN PB_T50_RS | overall accuracy: 90.01 -> 89.49 | Hurt |
| use DAPT instead of full fine-tuning | ACT · ScanObjectNN PB_T50_RS | overall accuracy: 88.21 -> 87.54 | Hurt |
| use Point-PEFT instead of full fine-tuning | ACT · ScanObjectNN OBJ_BG | overall accuracy: 93.29 -> 90.36 | Hurt |
| add voting at test time | PointGST on PointGPT-L · ScanObjectNN OBJ_BG | overall accuracy: 98.97 -> 99.48 | Improved |
| PointGST with strong rotation augmentation vs Point-PEFT | Point-MAE · ScanObjectNN PB_T50_RS | overall accuracy: 89.1 -> 89.3 | Improved |
| train on only 2% of ScanObjectNN PB_T50_RS data with PointGST vs other PEFT | PointGPT-L · ScanObjectNN PB_T50_RS few-shot | overall accuracy: 81.96% outperforming IDPT, Point-PEFT, and DAPT by 1.74%, 4.34%, and 4.51% | Improved |
| use PointGST instead of DAPT | ACT · S3DIS Area 5 | mIoU: 54.5 -> 57.4 | Improved |
| use PointGST instead of linear probing | Point-BERT · S3DIS Area 5 | mIoU: 53.9 -> 58.4 | Improved |
| use PointGST instead of full fine-tuning | PointGPT-L · ShapeNetPart | class mIoU: 84.80 -> 83.91 | Hurt |
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.
- Point-PEFT — Point-PEFT: Parameter-Efficient Fine-Tuning for 3D Pre-trained Models (AAAI 2024) · 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
- PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning (CVPR 2025 · similarity 0.88)
- Point-PEFT: Parameter-Efficient Fine-Tuning for 3D Pre-trained Models (AAAI 2024 · similarity 0.87)
- Dynamic Adapter Meets Prompt Tuning: Parameter-Efficient Transfer Learning for Point Cloud Analysis (CVPR 2024 · similarity 0.87)
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.85)
- PointPatchMix: Point Cloud Mixing with Patch Scoring (AAAI 2024 · similarity 0.85)
- PointRWKV: Efficient RWKV-Like Model for Hierarchical Point Cloud Learning (AAAI 2025 · 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.
