On the robustness of semantic segmentation models to adversarial attacks
Deep Neural Networks (DNNs) have demonstrated exceptional performance on most recognition tasks such as image classification and segmentation. However, they have also been shown to be vulnerable to adversarial examples. This phenomenon has recently attracted a lot of attention but it has not been ex...
| Main Authors: | , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
Institute of Electrical and Electronics Engineers
2019
|
| _version_ | 1826293123423141888 |
|---|---|
| author | Arnab, A Miksik, O Torr, PHS |
| author_facet | Arnab, A Miksik, O Torr, PHS |
| author_sort | Arnab, A |
| collection | OXFORD |
| description | Deep Neural Networks (DNNs) have demonstrated exceptional performance on most recognition tasks such as image classification and segmentation. However, they have also been shown to be vulnerable to adversarial examples. This phenomenon has recently attracted a lot of attention but it has not been extensively studied on multiple, large-scale datasets and structured prediction tasks such as semantic segmentation which often require more specialised networks with additional components such as CRFs, dilated convolutions, skip-connections and multiscale processing. In this paper, we present what to our knowledge is the first rigorous evaluation of adversarial attacks on modern semantic segmentation models, using two large-scale datasets. We analyse the effect of different network architectures, model capacity and multiscale processing, and show that many observations made on the task of classification do not always transfer to this more complex task. Furthermore, we show how mean-field inference in deep structured models, multiscale processing (and more generally, input transformations) naturally implement recently proposed adversarial defenses. Our observations will aid future efforts in understanding and defending against adversarial examples. Moreover, in the shorter term, we show how to effectively benchmark robustness and show which segmentation models should currently be preferred in safety-critical applications due to their inherent robustness. |
| first_indexed | 2024-03-07T03:25:14Z |
| format | Journal article |
| id | oxford-uuid:b8cf8a4d-4005-49ca-b7a0-22cc03eb5469 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T03:25:14Z |
| publishDate | 2019 |
| publisher | Institute of Electrical and Electronics Engineers |
| record_format | dspace |
| spelling | oxford-uuid:b8cf8a4d-4005-49ca-b7a0-22cc03eb54692022-03-27T04:58:32ZOn the robustness of semantic segmentation models to adversarial attacksJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:b8cf8a4d-4005-49ca-b7a0-22cc03eb5469EnglishSymplectic ElementsInstitute of Electrical and Electronics Engineers2019Arnab, AMiksik, OTorr, PHSDeep Neural Networks (DNNs) have demonstrated exceptional performance on most recognition tasks such as image classification and segmentation. However, they have also been shown to be vulnerable to adversarial examples. This phenomenon has recently attracted a lot of attention but it has not been extensively studied on multiple, large-scale datasets and structured prediction tasks such as semantic segmentation which often require more specialised networks with additional components such as CRFs, dilated convolutions, skip-connections and multiscale processing. In this paper, we present what to our knowledge is the first rigorous evaluation of adversarial attacks on modern semantic segmentation models, using two large-scale datasets. We analyse the effect of different network architectures, model capacity and multiscale processing, and show that many observations made on the task of classification do not always transfer to this more complex task. Furthermore, we show how mean-field inference in deep structured models, multiscale processing (and more generally, input transformations) naturally implement recently proposed adversarial defenses. Our observations will aid future efforts in understanding and defending against adversarial examples. Moreover, in the shorter term, we show how to effectively benchmark robustness and show which segmentation models should currently be preferred in safety-critical applications due to their inherent robustness. |
| spellingShingle | Arnab, A Miksik, O Torr, PHS On the robustness of semantic segmentation models to adversarial attacks |
| title | On the robustness of semantic segmentation models to adversarial attacks |
| title_full | On the robustness of semantic segmentation models to adversarial attacks |
| title_fullStr | On the robustness of semantic segmentation models to adversarial attacks |
| title_full_unstemmed | On the robustness of semantic segmentation models to adversarial attacks |
| title_short | On the robustness of semantic segmentation models to adversarial attacks |
| title_sort | on the robustness of semantic segmentation models to adversarial attacks |
| work_keys_str_mv | AT arnaba ontherobustnessofsemanticsegmentationmodelstoadversarialattacks AT miksiko ontherobustnessofsemanticsegmentationmodelstoadversarialattacks AT torrphs ontherobustnessofsemanticsegmentationmodelstoadversarialattacks |