Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Instance-aware Dynamic Prompt Tuning for Pre-trained Point Cloud Models

About

Pre-trained point cloud models have found extensive applications in 3D understanding tasks like object classification and part segmentation. However, the prevailing strategy of full fine-tuning in downstream tasks leads to large per-task storage overhead for model parameters, which limits the efficiency when applying large-scale pre-trained models. Inspired by the recent success of visual prompt tuning (VPT), this paper attempts to explore prompt tuning on pre-trained point cloud models, to pursue an elegant balance between performance and parameter efficiency. We find while instance-agnostic static prompting, e.g. VPT, shows some efficacy in downstream transfer, it is vulnerable to the distribution diversity caused by various types of noises in real-world point cloud data. To conquer this limitation, we propose a novel Instance-aware Dynamic Prompt Tuning (IDPT) strategy for pre-trained point cloud models. The essence of IDPT is to develop a dynamic prompt generation module to perceive semantic prior features of each point cloud instance and generate adaptive prompt tokens to enhance the model's robustness. Notably, extensive experiments demonstrate that IDPT outperforms full fine-tuning in most tasks with a mere 7% of the trainable parameters, providing a promising solution to parameter-efficient learning for pre-trained point cloud models. Code is available at \url{https://github.com/zyh16143998882/ICCV23-IDPT}.

Yaohua Zha, Jinpeng Wang, Tao Dai, Bin Chen, Zhi Wang, Shu-Tao Xia• 2023

Related benchmarks

TaskDatasetResultRank
Semantic segmentationS3DIS (Area 5)
mIOU50.5
907
Part SegmentationShapeNetPart (test)
mIoU (Inst.)85.9
312
Part SegmentationShapeNetPart
mIoU (Instance)85.9
246
Point Cloud ClassificationModelNet40 (test)
Accuracy93.4
229
Object ClassificationScanObjectNN OBJ_BG
Accuracy98.11
223
Object ClassificationScanObjectNN PB_T50_RS
Accuracy92.99
195
Object ClassificationModelNet40 (test)
Accuracy94.4
180
Object ClassificationScanObjectNN OBJ_ONLY
Overall Accuracy96.04
166
3D Semantic SegmentationScanNet (val)
mIoU72.6
144
Few-shot classificationModelNet40 10-way 20-shot
Accuracy95.5
105
Showing 10 of 55 rows

Other info

Code

Follow for update