{"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/two-stage-coarse-to-fine-image-anomaly","title":"Two-stage coarse-to-fine image anomaly segmentation and detection model","arxiv_id":null,"date":"2023-09-19","proceeding":"Image and Vision Computing 2023 9","authors":["Rizwan Ali Shah","Odilbek Urmonov","HyungWon Kim"],"abstract":"Existing Convolutional Neural Network (CNN) based anomaly detection and segmentation approaches are overly\r\nsensitive or not sensitive enough to noise, resulting in anomaly patterns, partially detected in the testing stage.\r\nThe previous methods may also differentiate normal and abnormal images, but they cannot identify the location\r\nof anomaly presented in test images with high accuracy. To address this issue, we propose a two-stage CNN\r\nmodel for coarse-to-fine anomaly segmentation and detection called (TASAD). In both stages of TASAD, we train\r\nour model on a mixture of normal and abnormal training samples. The abnormal images are obtained by insert-\r\ning pseudo-anomaly patterns that are automatically generated from anomaly source images. We use a novel and\r\nsophisticated anomaly insertion technique to generate various anomalous samples. In the first stage, we design a\r\ncoarse anomaly segmentation (CAS) model that takes a whole image as an input, while in the second stage, we\r\ntrain a fine anomaly segmentation (FAS) model on image patches. FAS model improves detection and segmenta-\r\ntion performance by refining anomaly patterns partially detected by CAS model. We train our framework on\r\nMVTec dataset and compare it with state-of-the-art (SOTA) methods. The proposed architecture leads to a com-\r\npact model size – four times smaller than the SOTA method, while exhibiting better pixel-level accuracy. TASAD\r\ncan also be applied to SOTAs to further improve their anomaly detection performance. Our experiments demon-\r\nstrate that when applied to the latest SOTAs, TASAD improves the average precision (AP) performance of previ-\r\nous methods by 6.2%. For reproducibility of the results, code is provided at https://github.com/RizwanAliQau/\r\ntasad.git.","url_abs":"https://authors.elsevier.com/c/1hnMnxnVKBFC-","url_pdf":"https://authors.elsevier.com/c/1hnMnxnVKBFC-","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":"two-stage-coarse-to-fine-image-anomaly","repo_url":"https://github.com/RizwanAliQau/tasad","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"anomaly-detection","task_name":"Anomaly Detection"},{"task_slug":"anomaly-segmentation","task_name":"Anomaly Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"}],"methods":[{"method_slug":"dcnn","method_name":"DCNN"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/anomaly-detection-on-mvtec-ad","task":"Anomaly Detection","dataset":"MVTec AD","model":"TASAD","rank_in_archive_order":67,"of":148,"metrics":{"Detection AUROC":"98.0","FPS":"142","Segmentation AP":"74.6","Segmentation AUROC":"98.2"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}