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

Monocular Depth Estimation (MDE) is a critical component in applications suchas autonomous driving. There are various attacks against MDE networks. Theseattacks, especially the physical ones, pose a great threat to the security ofsuch systems. Traditional adversarial training method requires ground-truthlabels hence cannot be directly applied to self-supervised MDE that does nothave ground-truth depth. Some self-supervised model hardening techniques (e.g.,contrastive learning) ignore the domain knowledge of MDE and can hardly achieveoptimal performance. In this work, we propose a novel adversarial trainingmethod for self-supervised MDE models based on view synthesis without usingground-truth depth. We improve adversarial robustness against physical-worldattacks using L0-norm-bounded perturbation in training. We compare our methodwith supervised learning based and contrastive learning based methods that aretailored for MDE. Results on two representative MDE networks show that weachieve better robustness against various adversarial attacks with nearly nobenign performance degradation.

Adversarial Training of Self-supervised Monocular Depth Estimation against Physical-World Attacks