Search papers, labs, and topics across Lattice.
This paper introduces the Point-Selection Fine-Tuning (PSFT) framework, which enhances the robustness of point cloud classification by selectively retaining minimally influential points to mitigate the impact of noisy and corrupted data. The method employs a prompt generation branch to inject layer-wise prompt tokens into a frozen backbone, coupled with a lightweight feature filter to refine representations before prediction. Experimental results demonstrate that PSFT significantly reduces corruption error across various 3D pre-trained models, achieving state-of-the-art performance on benchmark datasets like ModelNet-C and ScanObjectNN-C.
By intelligently filtering out noise, PSFT achieves superior robustness in point cloud classification, outperforming traditional fine-tuning methods under challenging conditions.
Noisy and corrupted points can substantially degrade point cloud recognition performance, especially under challenging corruption settings. In particular, full fine-tuning of 3D pre-trained models may amplify the influence of outliers and overwrite robustness priors learned during pre-training, while naive parameter-efficient adaptation remains sensitive to corrupted tokens. To address this issue, we propose PSFT, a point-selection fine-tuning framework that improves robustness while remaining parameter-efficient. PSFT first estimates point-wise influence from pre-pooling features and adaptively retains minimally influential points to suppress outliers. Based on the selected subset, a prompt generation branch predicts layer-wise prompt tokens and injects them into a frozen backbone for lightweight downstream adaptation. To further mitigate residual noise after selection, we append a lightweight feature filter with bottleneck MLP transformation and Beta-gated residual blending to refine patch-token representations before prediction. Extensive experiments show that PSFT consistently reduces corruption error on ModelNet-C and ModelNet40-C across all tested 3D pre-trained backbones, while achieving the strongest ScanObjectNN-C results with ULIP-2 and Uni3D-B among the evaluated tuning strategies. Our implementation can be found at https://github.com/CVChMA/PSFT/tree/master.
