{"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/imperfect-segmentation-labels-how-much-do","title":"Imperfect Segmentation Labels: How Much Do They Matter?","arxiv_id":"1806.04618","date":"2018-06-12","proceeding":null,"authors":["Nicholas Heller","Joshua Dean","Nikolaos Papanikolopoulos"],"abstract":"Labeled datasets for semantic segmentation are imperfect, especially in\nmedical imaging where borders are often subtle or ill-defined. Little work has\nbeen done to analyze the effect that label errors have on the performance of\nsegmentation methodologies. Here we present a large-scale study of model\nperformance in the presence of varying types and degrees of error in training\ndata. We trained U-Net, SegNet, and FCN32 several times for liver segmentation\nwith 10 different modes of ground-truth perturbation. Our results show that for\neach architecture, performance steadily declines with boundary-localized\nerrors, however, U-Net was significantly more robust to jagged boundary errors\nthan the other architectures. We also found that each architecture was very\nrobust to non-boundary-localized errors, suggesting that boundary-localized\nerrors are fundamentally different and more challenging problem than random\nlabel errors in a classification setting.","url_abs":"http://arxiv.org/abs/1806.04618v3","url_pdf":"http://arxiv.org/pdf/1806.04618v3.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":"imperfect-segmentation-labels-how-much-do","repo_url":"https://github.com/neheller/labels18","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"liver-segmentation","task_name":"Liver Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"segnet","method_name":"SegNet"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"u-net","method_name":"U-Net"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}