{"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/deep-fusion-network-for-image-completion","title":"Deep Fusion Network for Image Completion","arxiv_id":"1904.08060","date":"2019-04-17","proceeding":null,"authors":["Xin Hong","Pengfei Xiong","Renhe Ji","Haoqiang Fan"],"abstract":"Deep image completion usually fails to harmonically blend the restored image\ninto existing content, especially in the boundary area. This paper handles with\nthis problem from a new perspective of creating a smooth transition and\nproposes a concise Deep Fusion Network (DFNet). Firstly, a fusion block is\nintroduced to generate a flexible alpha composition map for combining known and\nunknown regions. The fusion block not only provides a smooth fusion between\nrestored and existing content, but also provides an attention map to make\nnetwork focus more on the unknown pixels. In this way, it builds a bridge for\nstructural and texture information, so that information can be naturally\npropagated from known region into completion. Furthermore, fusion blocks are\nembedded into several decoder layers of the network. Accompanied by the\nadjustable loss constraints on each layer, more accurate structure information\nare achieved. We qualitatively and quantitatively compare our method with other\nstate-of-the-art methods on Places2 and CelebA datasets. The results show the\nsuperior performance of DFNet, especially in the aspects of harmonious texture\ntransition, texture detail and semantic structural consistency. Our source code\nwill be avaiable at: \\url{https://github.com/hughplay/DFNet}","url_abs":"http://arxiv.org/abs/1904.08060v1","url_pdf":"http://arxiv.org/pdf/1904.08060v1.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":"deep-fusion-network-for-image-completion","repo_url":"https://github.com/researchmm/PEN-Net-for-Inpainting","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"deep-fusion-network-for-image-completion","repo_url":"https://github.com/hughplay/DFNet","is_official":0,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"deep-fusion-network-for-image-completion","repo_url":"https://github.com/iankuoli/DFNet_TF2","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"deep-fusion-network-for-image-completion","repo_url":"https://github.com/zphang/saliency_investigation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"deep-fusion-network-for-image-completion","repo_url":"https://github.com/deepcodebase/inpaint","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"image-inpainting","task_name":"Image Inpainting"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.08060","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}