{"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/dynamic-feature-fusion-for-semantic-edge","title":"Dynamic Feature Fusion for Semantic Edge Detection","arxiv_id":"1902.09104","date":"2019-02-25","proceeding":null,"authors":["Yuan Hu","Yunpeng Chen","Xiang Li","Jiashi Feng"],"abstract":"Features from multiple scales can greatly benefit the semantic edge detection\ntask if they are well fused. However, the prevalent semantic edge detection\nmethods apply a fixed weight fusion strategy where images with different\nsemantics are forced to share the same weights, resulting in universal fusion\nweights for all images and locations regardless of their different semantics or\nlocal context. In this work, we propose a novel dynamic feature fusion strategy\nthat assigns different fusion weights for different input images and locations\nadaptively. This is achieved by a proposed weight learner to infer proper\nfusion weights over multi-level features for each location of the feature map,\nconditioned on the specific input. In this way, the heterogeneity in\ncontributions made by different locations of feature maps and input images can\nbe better considered and thus help produce more accurate and sharper edge\npredictions. We show that our model with the novel dynamic feature fusion is\nsuperior to fixed weight fusion and also the na\\\"ive location-invariant weight\nfusion methods, via comprehensive experiments on benchmarks Cityscapes and SBD.\nIn particular, our method outperforms all existing well established methods and\nachieves new state-of-the-art.","url_abs":"http://arxiv.org/abs/1902.09104v1","url_pdf":"http://arxiv.org/pdf/1902.09104v1.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":"dynamic-feature-fusion-for-semantic-edge","repo_url":"https://github.com/Lavender105/DFF","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"edge-detection","task_name":"Edge Detection"}],"methods":[{"method_slug":"sbd","method_name":"Spatial Broadcast Decoder"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1902.09104","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}