English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Notwithstanding the prominent performance achieved in various applications,point cloud recognition models have often suffered from natural corruptions andadversarial perturbations. In this paper, we delve into boosting the generalrobustness of point cloud recognition models and propose Point-CloudContrastive Adversarial Training (PointCAT). The main intuition of PointCAT isencouraging the target recognition model to narrow the decision gap betweenclean point clouds and corrupted point clouds. Specifically, we leverage asupervised contrastive loss to facilitate the alignment and uniformity of thehypersphere features extracted by the recognition model, and design a pair ofcentralizing losses with the dynamic prototype guidance to avoid these featuresdeviating from their belonging category clusters. To provide the morechallenging corrupted point clouds, we adversarially train a noise generatoralong with the recognition model from the scratch, instead of usinggradient-based attack as the inner loop like previous adversarial trainingmethods. Comprehensive experiments show that the proposed PointCAT outperformsthe baseline methods and dramatically boosts the robustness of different pointcloud recognition models, under a variety of corruptions including isotropicpoint noises, the LiDAR simulated noises, random point dropping and adversarialperturbations.

PointCAT: Contrastive Adversarial Training for Robust Point Cloud Recognition