{"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/polarized-self-attention-towards-high-quality-1","title":"Polarized Self-Attention: Towards High-quality Pixel-wise Regression","arxiv_id":"2107.00782","date":"2021-07-02","proceeding":"arXiv preprint 2021 7","authors":["Huajun Liu","Fuqiang Liu","Xinyi Fan","Dong Huang"],"abstract":"Pixel-wise regression is probably the most common problem in fine-grained computer vision tasks, such as estimating keypoint heatmaps and segmentation masks. These regression problems are very challenging particularly because they require, at low computation overheads, modeling long-range dependencies on high-resolution inputs/outputs to estimate the highly nonlinear pixel-wise semantics. While attention mechanisms in Deep Convolutional Neural Networks(DCNNs) has become popular for boosting long-range dependencies, element-specific attention, such as Nonlocal blocks, is highly complex and noise-sensitive to learn, and most of simplified attention hybrids try to reach the best compromise among multiple types of tasks. In this paper, we present the Polarized Self-Attention(PSA) block that incorporates two critical designs towards high-quality pixel-wise regression: (1) Polarized filtering: keeping high internal resolution in both channel and spatial attention computation while completely collapsing input tensors along their counterpart dimensions. (2) Enhancement: composing non-linearity that directly fits the output distribution of typical fine-grained regression, such as the 2D Gaussian distribution (keypoint heatmaps), or the 2D Binormial distribution (binary segmentation masks). PSA appears to have exhausted the representation capacity within its channel-only and spatial-only branches, such that there is only marginal metric differences between its sequential and parallel layouts. Experimental results show that PSA boosts standard baselines by $2-4$ points, and boosts state-of-the-arts by $1-2$ points on 2D pose estimation and semantic segmentation benchmarks.","url_abs":"https://arxiv.org/abs/2107.00782v2","url_pdf":"https://arxiv.org/pdf/2107.00782v2.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":"polarized-self-attention-towards-high-quality-1","repo_url":"https://github.com/DeLightCMU/PSA","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"polarized-self-attention-towards-high-quality-1","repo_url":"https://github.com/sithu31296/pose-estimation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"polarized-self-attention-towards-high-quality-1","repo_url":"https://github.com/sithu31296/semantic-segmentation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"polarized-self-attention-towards-high-quality-1","repo_url":"https://github.com/PaddlePaddle/PaddleSeg","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null},{"paper_slug":"polarized-self-attention-towards-high-quality-1","repo_url":"https://github.com/mindspore-courses/External-Attention-MindSpore/blob/main/model/attention/PolarizedSelfAttention.py","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"polarized-self-attention-towards-high-quality-1","repo_url":"https://github.com/xmu-xiaoma666/External-Attention-pytorch/blob/master/attention/PolarizedSelfAttention.py","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"2d-pose-estimation","task_name":"2D Pose Estimation"},{"task_slug":"keypoint-detection","task_name":"Keypoint Detection"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"high","task_name":"Vocal Bursts Intensity Prediction"},{"task_slug":"regression-1","task_name":"regression"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/keypoint-detection-on-coco","task":"Keypoint Detection","dataset":"COCO (Common Objects in Context)","model":"UDP-Pose-PSA(384x288)","rank_in_archive_order":20,"of":24,"metrics":{"Validation AP":"79.5"},"uses_additional_data":false},{"leaderboard":"/sota/pose-estimation-on-coco-test-dev","task":"Pose Estimation","dataset":"COCO test-dev","model":"UDP-Pose-PSA(384x288)","rank_in_archive_order":3,"of":47,"metrics":{"AP":"79.5","AP50":"93.6","AP75":"85.9","APL":"84.3","APM":"76.3","AR":"81.9"},"uses_additional_data":false},{"leaderboard":"/sota/pose-estimation-on-coco-test-dev","task":"Pose Estimation","dataset":"COCO test-dev","model":"UDP-Pose-PSA(256x192)","rank_in_archive_order":7,"of":47,"metrics":{"AP":"78.9","AP50":"93.6","AP75":"85.8","APL":"83.6","APM":"76.1","AR":"81.4"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-cityscapes-val","task":"Semantic Segmentation","dataset":"Cityscapes val","model":"HRNetV2-OCR+PSA","rank_in_archive_order":5,"of":99,"metrics":{"mIoU":"86.93"},"uses_additional_data":true}],"syntology":{"syntology_url":"https://syntology.ai/paper/2107.00782","atlas_url":"https://app.syntology.ai/?focus=2107.00782","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}