{"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/starmap-for-category-agnostic-keypoint-and","title":"StarMap for Category-Agnostic Keypoint and Viewpoint Estimation","arxiv_id":"1803.09331","date":"2018-03-25","proceeding":"ECCV 2018 9","authors":["Xingyi Zhou","Arjun Karpur","Linjie Luo","Qi-Xing Huang"],"abstract":"Semantic keypoints provide concise abstractions for a variety of visual\nunderstanding tasks. Existing methods define semantic keypoints separately for\neach category with a fixed number of semantic labels in fixed indices. As a\nresult, this keypoint representation is in-feasible when objects have a varying\nnumber of parts, e.g. chairs with varying number of legs. We propose a\ncategory-agnostic keypoint representation, which combines a multi-peak heatmap\n(StarMap) for all the keypoints and their corresponding features as 3D\nlocations in the canonical viewpoint (CanViewFeature) defined for each\ninstance. Our intuition is that the 3D locations of the keypoints in canonical\nobject views contain rich semantic and compositional information. Using our\nflexible representation, we demonstrate competitive performance in keypoint\ndetection and localization compared to category-specific state-of-the-art\nmethods. Moreover, we show that when augmented with an additional depth channel\n(DepthMap) to lift the 2D keypoints to 3D, our representation can achieve\nstate-of-the-art results in viewpoint estimation. Finally, we show that our\ncategory-agnostic keypoint representation can be generalized to novel\ncategories.","url_abs":"http://arxiv.org/abs/1803.09331v2","url_pdf":"http://arxiv.org/pdf/1803.09331v2.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":"starmap-for-category-agnostic-keypoint-and","repo_url":"https://github.com/xingyizhou/StarMap","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"keypoint-detection","task_name":"Keypoint Detection"},{"task_slug":"viewpoint-estimation","task_name":"Viewpoint Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/keypoint-detection-on-pascal3d","task":"Keypoint Detection","dataset":"Pascal3D+","model":"StarMap","rank_in_archive_order":2,"of":4,"metrics":{"Mean PCK":"78.6"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1803.09331","atlas_url":"https://app.syntology.ai/?focus=1803.09331","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}