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

Stereo matching has recently witnessed remarkable progress using Deep NeuralNetworks (DNNs). But, how robust are they? Although it has been well-known thatDNNs often suffer from adversarial vulnerability with a catastrophic drop inperformance, the situation is even worse in stereo matching. This paper firstshows that a type of weak white-box attacks can overwhelm state-of-the-artmethods. The attack is learned by a proposed stereo-constrained projectedgradient descent (PGD) method in stereo matching. This observation raisesserious concerns for the deployment of DNN-based stereo matching. Parallel tothe adversarial vulnerability, DNN-based stereo matching is typically trainedunder the so-called simulation to reality pipeline, and thus domaingeneralizability is an important problem. This paper proposes to rethink thelearnable DNN-based feature backbone towards adversarially-robust and domaingeneralizable stereo matching by completely removing it for matching. Inexperiments, the proposed method is tested in the SceneFlow dataset and theKITTI2015 benchmark, with promising results. We compute the matching costvolume using the classic multi-scale census transform (i.e., local binarypattern) of the raw input stereo images, followed by a stacked Hourglass headsub-network solving the matching problem. It significantly improves theadversarial robustness, while retaining accuracy performance comparable tostate-of-the-art methods. It also shows better generalizability from simulation(SceneFlow) to real (KITTI) datasets when no fine-tuning is used.

Towards Adversarially Robust and Domain Generalizable Stereo Matching by Rethinking DNN Feature Backbones