{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/robustness-of-3d-deep-learning-in-an","title":"Robustness of 3D Deep Learning in an Adversarial Setting","arxiv_id":"1904.00923","date":"2019-04-01","proceeding":"CVPR 2019 6","authors":["Matthew Wicker","Marta Kwiatkowska"],"abstract":"Understanding the spatial arrangement and nature of real-world objects is of\nparamount importance to many complex engineering tasks, including autonomous\nnavigation. Deep learning has revolutionized state-of-the-art performance for\ntasks in 3D environments; however, relatively little is known about the\nrobustness of these approaches in an adversarial setting. The lack of\ncomprehensive analysis makes it difficult to justify deployment of 3D deep\nlearning models in real-world, safety-critical applications. In this work, we\ndevelop an algorithm for analysis of pointwise robustness of neural networks\nthat operate on 3D data. We show that current approaches presented for\nunderstanding the resilience of state-of-the-art models vastly overestimate\ntheir robustness. We then use our algorithm to evaluate an array of\nstate-of-the-art models in order to demonstrate their vulnerability to\nocclusion attacks. We show that, in the worst case, these networks can be\nreduced to 0% classification accuracy after the occlusion of at most 6.5% of\nthe occupied input space.","url_abs":"http://arxiv.org/abs/1904.00923v1","url_pdf":"http://arxiv.org/pdf/1904.00923v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"robustness-of-3d-deep-learning-in-an","repo_url":"https://github.com/matthewwicker/IterativeSalienceOcclusion","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"autonomous-navigation","task_name":"Autonomous Navigation"},{"task_slug":"deep-learning","task_name":"Deep Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1904.00923","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}