{"url":"/method/gsop-net","slug":"gsop-net","name":"GSoP-Net","full_name":"Global second-order pooling convolutional networks","full_name_withheld":false,"description_markdown":"A Gsop block has a squeeze module and an excitation module, and uses a second-order pooling to model high-order statistics while gathering global information.\r\nIn the squeeze module, a GSoP block firstly reduces the number of channels from $c$ to $c'$ ($c' < c$) using a $1 \\times 1$ convolution,  then  computes a $c' \\times c'$ covariance matrix for the different channels to obtain their correlation.  Next, row-wise normalization is performed on the covariance matrix.  Each $(i, j)$ in the normalized covariance matrix explicitly relates channel $i$ to channel $j$. \r\n\r\nIn the excitation module, a GSoP block performs row-wise convolution to  maintain structural information and output a vector. Then a fully-connected layer and a sigmoid function are applied  to get a $c$-dimensional attention vector. Finally, it multiplies the input features by the attention vector, as in an SE block. A GSoP block can be formulated as:\r\n\\begin{align}\r\n    s = F_\\text{gsop}(X, \\theta) & = \\sigma (W \\text{RC}(\\text{Cov}(\\text{Conv}(X))))\r\n\\end{align}\r\n\\begin{align}\r\n    Y & = s  X\r\n\\end{align}\r\nHere, $\\text{Conv}(\\cdot)$ reduces the number of channels,\r\n$\\text{Cov}(\\cdot)$ computes the covariance matrix and\r\n$\\text{RC}(\\cdot)$ means row-wise convolution.","description_state":"present","introduced_year":null,"introduced_by":{"title":"Global Second-order Pooling Convolutional Networks","paper":"/paper/global-second-order-pooling-convolutional","first_author":"Zilin Gao","n_authors":4,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/global-second-order-pooling-convolutional"},"source":{"url":"http://arxiv.org/abs/1811.12006v2","title":"Global Second-order Pooling Convolutional Networks","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":1,"archive_num_papers":1,"papers_newest_first":[{"paper":"/paper/global-second-order-pooling-convolutional","title":"Global Second-order Pooling Convolutional Networks","date":"2018-11-29","arxiv_id":"1811.12006","n_code_links":1,"syntology":{"ran":2,"of":3,"unverified":1,"pointer_only":1}}],"papers_shown":1,"tasks":[{"task":"/task/object-recognition","name":"Object Recognition","papers":1}],"tasks_shown":1,"n_tasks":1,"usage_by_year":[{"year":"2018","papers":1}],"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/gsop-net"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}