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/

Zendy Plus

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

Zendy Tools

Zendy Open

Deep neural networks (DNNs) have been proven extremely susceptible toadversarial examples, which raises special safety-critical concerns forDNN-based autonomous driving stacks (i.e., 3D object detection). Although thereare extensive works on image-level attacks, most are restricted to 2D pixelspaces, and such attacks are not always physically realistic in our 3D world.Here we present Adv3D, the first exploration of modeling adversarial examplesas Neural Radiance Fields (NeRFs). Advances in NeRF provide photorealisticappearances and 3D accurate generation, yielding a more realistic andrealizable adversarial example. We train our adversarial NeRF by minimizing thesurrounding objects' confidence predicted by 3D detectors on the training set.Then we evaluate Adv3D on the unseen validation set and show that it can causea large performance reduction when rendering NeRF in any sampled pose. Togenerate physically realizable adversarial examples, we propose primitive-awaresampling and semantic-guided regularization that enable 3D patch attacks withcamouflage adversarial texture. Experimental results demonstrate that thetrained adversarial NeRF generalizes well to different poses, scenes, and 3Ddetectors. Finally, we provide a defense method to our attacks that involvesadversarial training through data augmentation. Project page:https://len-li.github.io/adv3d-web

Adv3D: Generating 3D Adversarial Examples for 3D Object Detection in Driving Scenarios with NeRF