{"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/fuseformer-fusing-fine-grained-information-in","title":"FuseFormer: Fusing Fine-Grained Information in Transformers for Video Inpainting","arxiv_id":"2109.02974","date":"2021-09-07","proceeding":"ICCV 2021 10","authors":["Rui Liu","Hanming Deng","Yangyi Huang","Xiaoyu Shi","Lewei Lu","Wenxiu Sun","Xiaogang Wang","Jifeng Dai","Hongsheng Li"],"abstract":"Transformer, as a strong and flexible architecture for modelling long-range relations, has been widely explored in vision tasks. However, when used in video inpainting that requires fine-grained representation, existed method still suffers from yielding blurry edges in detail due to the hard patch splitting. Here we aim to tackle this problem by proposing FuseFormer, a Transformer model designed for video inpainting via fine-grained feature fusion based on novel Soft Split and Soft Composition operations. The soft split divides feature map into many patches with given overlapping interval. On the contrary, the soft composition operates by stitching different patches into a whole feature map where pixels in overlapping regions are summed up. These two modules are first used in tokenization before Transformer layers and de-tokenization after Transformer layers, for effective mapping between tokens and features. Therefore, sub-patch level information interaction is enabled for more effective feature propagation between neighboring patches, resulting in synthesizing vivid content for hole regions in videos. Moreover, in FuseFormer, we elaborately insert the soft composition and soft split into the feed-forward network, enabling the 1D linear layers to have the capability of modelling 2D structure. And, the sub-patch level feature fusion ability is further enhanced. In both quantitative and qualitative evaluations, our proposed FuseFormer surpasses state-of-the-art methods. We also conduct detailed analysis to examine its superiority.","url_abs":"https://arxiv.org/abs/2109.02974v1","url_pdf":"https://arxiv.org/pdf/2109.02974v1.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":"fuseformer-fusing-fine-grained-information-in","repo_url":"https://github.com/ruiliu-ai/fuseformer","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"seeing-beyond-the-visible","task_name":"Seeing Beyond the Visible"},{"task_slug":"video-inpainting","task_name":"Video Inpainting"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"fuseformer","method_name":"FuseFormer"},{"method_slug":"fuseformer-block","method_name":"FuseFormer Block"},{"method_slug":"pixel-prediction","method_name":"Inpainting"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"soft-split-and-soft-composition","method_name":"Soft Split and Soft Composition"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[{"slug":"fuseformer-block","name":"FuseFormer Block","full_name":"FuseFormer Block"},{"slug":"soft-split-and-soft-composition","name":"Soft Split and Soft Composition","full_name":"Soft Split and Soft Composition"}],"results":[{"leaderboard":"/sota/seeing-beyond-the-visible-on-kitti360-ex","task":"Seeing Beyond the Visible","dataset":"KITTI360-EX","model":"FuseFormer","rank_in_archive_order":4,"of":7,"metrics":{"Average PSNR":"18.91"},"uses_additional_data":false},{"leaderboard":"/sota/video-inpainting-on-davis","task":"Video Inpainting","dataset":"DAVIS","model":"FuseFormer","rank_in_archive_order":3,"of":11,"metrics":{"Ewarp":"0.1362","PSNR":"32.54","SSIM":"0.9700","VFID":"0.138"},"uses_additional_data":false},{"leaderboard":"/sota/video-inpainting-on-hqvi-240p","task":"Video Inpainting","dataset":"HQVI (240p)","model":"FuseFormer","rank_in_archive_order":6,"of":7,"metrics":{"LPIPS":"0.0498","PSNR":"29.92","SSIM":"0.9365","VFID":"0.2727"},"uses_additional_data":false},{"leaderboard":"/sota/video-inpainting-on-youtube-vos","task":"Video Inpainting","dataset":"YouTube-VOS 2018","model":"FuseFormer","rank_in_archive_order":4,"of":10,"metrics":{"Ewarp":"0.0900","PSNR":"33.29","SSIM":"0.9681","VFID":"0.053"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2109.02974","atlas_url":"https://app.syntology.ai/?focus=2109.02974","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2109.02974"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ruiliu-ai/fuseformer","reach":{"status":"ok"}},{"provenance":"deterministic:regex_extraction","url":"https://github.com/ruiliu-ai/FuseFormer","reach":{"status":"ok"}}],"summary":{"ran":10,"unverified":1},"by_repo_kind":{"official":{"samples":11,"ran":10,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":11,"samples":[{"code_sha256_prefix":"974fc45de281481c","entry":"AddPosEmb","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"974fc45de281481c"}},{"code_sha256_prefix":"a702c7ce4cea9802","entry":"Attention","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"a702c7ce4cea9802"}},{"code_sha256_prefix":"6a1c3869290d0d57","entry":"BaseNetwork","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"6a1c3869290d0d57"}},{"code_sha256_prefix":"5231e4794aac485a","entry":"Encoder","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"5231e4794aac485a"}},{"code_sha256_prefix":"3d843c10ed59c635","entry":"FusionFeedForward","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"3d843c10ed59c635"}},{"code_sha256_prefix":"634deecea407ff83","entry":"MultiHeadedAttention","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"634deecea407ff83"}},{"code_sha256_prefix":"b99e0f5e4538ded3","entry":"SoftComp","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"b99e0f5e4538ded3"}},{"code_sha256_prefix":"2e02b61d792f9198","entry":"SoftSplit","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"2e02b61d792f9198"}},{"code_sha256_prefix":"afa615aa2f60ae16","entry":"TransformerBlock","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"afa615aa2f60ae16"}},{"code_sha256_prefix":"e73d218a85e7f736","entry":"deconv","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"e73d218a85e7f736"}},{"code_sha256_prefix":"dbafafc1baa0fc1c","entry":"InpaintGenerator","repo":"ruiliu-ai/FuseFormer","repo_kind":"official","path":"model/fuseformer.py","file_url":"https://github.com/ruiliu-ai/FuseFormer/blob/HEAD/model/fuseformer.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"dbafafc1baa0fc1c"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}