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 network-based image classifications are vulnerable to adversarialperturbations. The image classifications can be easily fooled by addingartificial small and imperceptible perturbations to input images. As one of themost effective defense strategies, adversarial training was proposed to addressthe vulnerability of classification models, where the adversarial examples arecreated and injected into training data during training. The attack and defenseof classification models have been intensively studied in past years. Semanticsegmentation, as an extension of classifications, has also received greatattention recently. Recent work shows a large number of attack iterations arerequired to create effective adversarial examples to fool segmentation models.The observation makes both robustness evaluation and adversarial training onsegmentation models challenging. In this work, we propose an effective andefficient segmentation attack method, dubbed SegPGD. Besides, we provide aconvergence analysis to show the proposed SegPGD can create more effectiveadversarial examples than PGD under the same number of attack iterations.Furthermore, we propose to apply our SegPGD as the underlying attack method forsegmentation adversarial training. Since SegPGD can create more effectiveadversarial examples, the adversarial training with our SegPGD can boost therobustness of segmentation models. Our proposals are also verified withexperiments on popular Segmentation model architectures and standardsegmentation datasets.

SegPGD: An Effective and Efficient Adversarial Attack for Evaluating and Boosting Segmentation Robustness