{"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/defectnet-multi-class-fault-detection-on","title":"DefectNET: multi-class fault detection on highly-imbalanced datasets","arxiv_id":"1904.00863","date":"2019-04-01","proceeding":null,"authors":["N. Anantrasirichai","David Bull"],"abstract":"As a data-driven method, the performance of deep convolutional neural\nnetworks (CNN) relies heavily on training data. The prediction results of\ntraditional networks give a bias toward larger classes, which tend to be the\nbackground in the semantic segmentation task. This becomes a major problem for\nfault detection, where the targets appear very small on the images and vary in\nboth types and sizes. In this paper we propose a new network architecture,\nDefectNet, that offers multi-class (including but not limited to) defect\ndetection on highly-imbalanced datasets. DefectNet consists of two parallel\npaths, which are a fully convolutional network and a dilated convolutional\nnetwork to detect large and small objects respectively. We propose a hybrid\nloss maximising the usefulness of a dice loss and a cross entropy loss, and we\nalso employ the leaky rectified linear unit (ReLU) to deal with rare occurrence\nof some targets in training batches. The prediction results show that our\nDefectNet outperforms state-of-the-art networks for detecting multi-class\ndefects with the average accuracy improvement of approximately 10% on a wind\nturbine.","url_abs":"http://arxiv.org/abs/1904.00863v2","url_pdf":"http://arxiv.org/pdf/1904.00863v2.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":"defectnet-multi-class-fault-detection-on","repo_url":"https://github.com/pui-nantheera/DefectNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"defect-detection","task_name":"Defect Detection"},{"task_slug":"fault-detection","task_name":"Fault Detection"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[{"method_slug":"dice-loss","method_name":"Dice Loss"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}