PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning
Song Wang, Xiaolu Liu, Lingdong Kong, Jianyun Xu, Chunyong Hu, Gongfan Fang, Wentong Li, Jianke Zhu
Abstract
Self-supervised representation learning for point cloud has demonstrated effectiveness in improving pre-trained model performance across diverse tasks. However, as pre-trained models grow in complexity, fully fine-tuning them for downstream applications demands substantial computational and storage resources. Parameter-efficient fine-tuning (PEFT) methods offer a promising solution to mitigate these resource requirements, yet most current approaches rely on complex adapter and prompt mechanisms that increase tunable parameters. In this paper, we propose PointLoRA, a simple yet effective method that combines low-rank adaptation (LoRA) with multi-scale token selection to efficiently fine-tune point cloud models. Our approach embeds LoRA layers within the most parameter-intensive components of point cloud transformers, reducing the need for tunable parameters while enhancing global feature capture. Additionally, multi-scale token selection extracts critical local information to serve as prompts for downstream fine-tuning, effectively complementing the global context captured by LoRA. The experimental results across various pre-trained models and three challenging public datasets demonstrate that our approach achieves competitive performance with only 3.43% of the trainable parameters, making it highly effective for resource-constrained applications. Source code is available at: https://github.com/songw-zju/PointLoRA.
BibTeX
@InProceedings{Wang_2025_CVPR,
author = {Wang, Song and Liu, Xiaolu and Kong, Lingdong and Xu, Jianyun and Hu, Chunyong and Fang, Gongfan and Li, Wentong and Zhu, Jianke and Wang, Xinchao},
title = {PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning},
booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)},
month = {June},
year = {2025},
pages = {6605-6615}
}