{"url":"/method/scse-1","slug":"scse-1","name":"scSE","full_name":"Spatial and Channel SE Blocks","full_name_withheld":false,"description_markdown":"To aggregate global spatial information,\r\nan SE block applies global pooling to the feature map.\r\nHowever, it ignores pixel-wise spatial information,\r\nwhich is important in dense prediction tasks.\r\nTherefore, Roy et al. proposed\r\nspatial and channel SE blocks (scSE). \r\nLike BAM, spatial SE blocks are used, complementing SE blocks, \r\nto provide spatial attention weights to focus on important regions.\r\n\r\nGiven the input feature map $X$, two parallel modules, spatial SE and channel SE, are applied to feature maps to encode spatial and channel information respectively. The channel SE module is an ordinary SE block, while the spatial SE module adopts $1\\times 1$ convolution for spatial squeezing. The outputs from the two modules are fused. The overall process can be written as\r\n\\begin{align}\r\n    s_c & = \\sigma (W_{2} \\delta (W_{1}\\text{GAP}(X)))\r\n\\end{align}\r\n\\begin{align}\r\n    X_\\text{chn} & = s_c  X \r\n\\end{align}\r\n\\begin{align}\r\n    s_s &= \\sigma(\\text{Conv}^{1\\times 1}(X))\r\n\\end{align}\r\n\\begin{align}\r\n    X_\\text{spa} & = s_s  X\r\n\\end{align}\r\n\\begin{align}\r\n    Y &= f(X_\\text{spa},X_\\text{chn})  \r\n\\end{align}\r\n\r\nwhere $f$ denotes the fusion function, which can be  maximum, addition, multiplication or concatenation. \r\n\r\nThe proposed scSE block combines channel and spatial attention to\r\nenhance features as well as \r\ncapturing pixel-wise spatial information.\r\nSegmentation tasks are greatly benefited as a result.\r\nThe integration of an scSE block in F-CNNs makes a consistent improvement in semantic segmentation at negligible extra cost.","description_state":"present","introduced_year":null,"introduced_by":{"title":"Recalibrating Fully Convolutional Networks with Spatial and Channel 'Squeeze & Excitation' Blocks","paper":"/paper/recalibrating-fully-convolutional-networks","first_author":"Abhijit Guha Roy","n_authors":3,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/recalibrating-fully-convolutional-networks"},"source":{"url":"http://arxiv.org/abs/1808.08127v1","title":"Recalibrating Fully Convolutional Networks with Spatial and Channel 'Squeeze & Excitation' Blocks","url_on_a_paper_host":true},"code_snippet_url":null,"code_snippet_url_on_a_code_host":false,"categories":[{"area":"General","area_id":"general","collection":"Attention Mechanisms","url":"/methods/category/attention-mechanisms","pwc_aliases":["attention-mechanisms-1"]}],"n_papers_tagged":4,"archive_num_papers":4,"papers_newest_first":[{"paper":null,"title":"Semi-supervised Learning for Segmentation of Bleeding Regions in Video Capsule Endoscopy","date":"2023-08-05","arxiv_id":"2308.02869","n_code_links":0,"syntology":null},{"paper":"/paper/fusegnet-a-deep-convolutional-neural-network","title":"FUSegNet: A Deep Convolutional Neural Network for Foot Ulcer Segmentation","date":"2023-05-04","arxiv_id":"2305.02961","n_code_links":1,"syntology":null},{"paper":null,"title":"Attention-based convolutional neural network for perfusion T2-weighted MR images preprocessing","date":"2023-03-04","arxiv_id":"2303.02518","n_code_links":0,"syntology":null},{"paper":"/paper/recalibrating-fully-convolutional-networks","title":"Recalibrating Fully Convolutional Networks with Spatial and Channel 'Squeeze & Excitation' Blocks","date":"2018-08-23","arxiv_id":"1808.08127","n_code_links":5,"syntology":{"ran":2,"of":8,"unverified":6,"pointer_only":0}}],"papers_shown":4,"tasks":[{"task":"/task/segmentation","name":"Segmentation","papers":2},{"task":"/task/anatomy","name":"Anatomy","papers":1},{"task":"/task/decoder","name":"Decoder","papers":1},{"task":"/task/diagnostic","name":"Diagnostic","papers":1},{"task":"/task/image-classification","name":"Image Classification","papers":1},{"task":"/task/semantic-segmentation","name":"Semantic Segmentation","papers":1},{"task":"/task/skull-stripping","name":"Skull Stripping","papers":1},{"task":"/task/image-classification","name":"image-classification","papers":1}],"tasks_shown":8,"n_tasks":8,"usage_by_year":[{"year":"2018","papers":1},{"year":"2023","papers":3}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/scse-1"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}