{"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/gcnet-non-local-networks-meet-squeeze","title":"GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond","arxiv_id":"1904.11492","date":"2019-04-25","proceeding":null,"authors":["Yue Cao","Jiarui Xu","Stephen Lin","Fangyun Wei","Han Hu"],"abstract":"The Non-Local Network (NLNet) presents a pioneering approach for capturing\nlong-range dependencies, via aggregating query-specific global context to each\nquery position. However, through a rigorous empirical analysis, we have found\nthat the global contexts modeled by non-local network are almost the same for\ndifferent query positions within an image. In this paper, we take advantage of\nthis finding to create a simplified network based on a query-independent\nformulation, which maintains the accuracy of NLNet but with significantly less\ncomputation. We further observe that this simplified design shares similar\nstructure with Squeeze-Excitation Network (SENet). Hence we unify them into a\nthree-step general framework for global context modeling. Within the general\nframework, we design a better instantiation, called the global context (GC)\nblock, which is lightweight and can effectively model the global context. The\nlightweight property allows us to apply it for multiple layers in a backbone\nnetwork to construct a global context network (GCNet), which generally\noutperforms both simplified NLNet and SENet on major benchmarks for various\nrecognition tasks. The code and configurations are released at\nhttps://github.com/xvjiarui/GCNet.","url_abs":"http://arxiv.org/abs/1904.11492v1","url_pdf":"http://arxiv.org/pdf/1904.11492v1.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":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/xvjiarui/GCNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/Oichii/DeepPulse-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/czero69/acomoeye-NN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/xggIoU/GCNet_global_context_module_tensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/zhusiling/GCNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/PaddlePaddle/PaddleSeg","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/open-mmlab/mmdetection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/open-mmlab/mmpose","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"gcnet-non-local-networks-meet-squeeze","repo_url":"https://github.com/open-mmlab/mmsegmentation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"instance-segmentation","task_name":"Instance Segmentation"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-recognition","task_name":"Object Recognition"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"bottleneck-residual-block","method_name":"Bottleneck Residual Block"},{"method_slug":"cascade-mask-r-cnn","method_name":"Cascade Mask R-CNN"},{"method_slug":"cascade-r-cnn","method_name":"Cascade R-CNN"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"cosine-annealing","method_name":"Cosine Annealing"},{"method_slug":"deformable-convolution","method_name":"Deformable Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"fpn","method_name":"FPN"},{"method_slug":"gcnet","method_name":"GCNet"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"global-context-block","method_name":"Global Context Block"},{"method_slug":"grouped-convolution","method_name":"Grouped Convolution"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-warmup-with-cosine-annealing","method_name":"Linear Warmup With Cosine Annealing"},{"method_slug":"mask-r-cnn","method_name":"Mask R-CNN"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"non-local-block","method_name":"Non-Local Block"},{"method_slug":"non-local-operation","method_name":"Non-Local Operation"},{"method_slug":"rpn","method_name":"RPN"},{"method_slug":"randomhorizontalflip","method_name":"Random Horizontal Flip"},{"method_slug":"random-resized-crop","method_name":"Random Resized Crop"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"resnext","method_name":"ResNeXt"},{"method_slug":"resnext-block","method_name":"ResNeXt Block"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"roi-align","method_name":"RoIAlign"},{"method_slug":"senet","method_name":"SENet"},{"method_slug":"sgd-with-momentum","method_name":"SGD with Momentum"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"squeeze-and-excitation-block","method_name":"Squeeze-and-Excitation Block"},{"method_slug":"step-decay","method_name":"Step Decay"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[],"methods_introduced":[{"slug":"gcnet","name":"GCNet","full_name":"GCNet"},{"slug":"global-context-block","name":"Global Context Block","full_name":"Global Context Block"}],"results":[{"leaderboard":"/sota/instance-segmentation-on-coco-minival","task":"Instance Segmentation","dataset":"COCO minival","model":"GCNet (ResNeXt-101 + DCN + cascade + GC r16)","rank_in_archive_order":70,"of":93,"metrics":{"mask AP":"40.9"},"uses_additional_data":false},{"leaderboard":"/sota/instance-segmentation-on-coco","task":"Instance Segmentation","dataset":"COCO test-dev","model":"GCNet (ResNeXt-101 + DCN + cascade + GC r16)","rank_in_archive_order":60,"of":112,"metrics":{"mask AP":"41.5%"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"GCNet (ResNeXt-101 + DCN + cascade + GC r16)","rank_in_archive_order":93,"of":220,"metrics":{"AP50":"66.9","AP75":"52.2","box AP":"47.9"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"GCnet (ResNet-50-FPN, GRoIE)","rank_in_archive_order":178,"of":220,"metrics":{"AP50":"62.4","AP75":"44","APL":"52.5","APM":"44.4","APS":"24.2","box AP":"40.3"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco","task":"Object Detection","dataset":"COCO test-dev","model":"GCNet (ResNeXt-101 + DCN + cascade + GC r4)","rank_in_archive_order":106,"of":225,"metrics":{"AP50":"67.6","AP75":"52.7","Operations per network pass":"54.8G","box mAP":"48.4"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-o","task":"Object Detection","dataset":"COCO-O","model":"GCNet\n(RX-101-32x4d-DCN)","rank_in_archive_order":25,"of":45,"metrics":{"Average mAP":"26.0","Effective Robustness":"4.38"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1904.11492","atlas_url":"https://app.syntology.ai/?focus=1904.11492","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}