{"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/superpixels-an-evaluation-of-the-state-of-the","title":"Superpixels: An Evaluation of the State-of-the-Art","arxiv_id":"1612.01601","date":"2016-12-06","proceeding":null,"authors":["David Stutz","Alexander Hermans","Bastian Leibe"],"abstract":"Superpixels group perceptually similar pixels to create visually meaningful\nentities while heavily reducing the number of primitives for subsequent\nprocessing steps. As of these properties, superpixel algorithms have received\nmuch attention since their naming in 2003. By today, publicly available\nsuperpixel algorithms have turned into standard tools in low-level vision. As\nsuch, and due to their quick adoption in a wide range of applications,\nappropriate benchmarks are crucial for algorithm selection and comparison.\nUntil now, the rapidly growing number of algorithms as well as varying\nexperimental setups hindered the development of a unifying benchmark. We\npresent a comprehensive evaluation of 28 state-of-the-art superpixel algorithms\nutilizing a benchmark focussing on fair comparison and designed to provide new\ninsights relevant for applications. To this end, we explicitly discuss\nparameter optimization and the importance of strictly enforcing connectivity.\nFurthermore, by extending well-known metrics, we are able to summarize\nalgorithm performance independent of the number of generated superpixels,\nthereby overcoming a major limitation of available benchmarks. Furthermore, we\ndiscuss runtime, robustness against noise, blur and affine transformations,\nimplementation details as well as aspects of visual quality. Finally, we\npresent an overall ranking of superpixel algorithms which redefines the\nstate-of-the-art and enables researchers to easily select appropriate\nalgorithms and the corresponding implementations which themselves are made\npublicly available as part of our benchmark at\ndavidstutz.de/projects/superpixel-benchmark/.","url_abs":"http://arxiv.org/abs/1612.01601v3","url_pdf":"http://arxiv.org/pdf/1612.01601v3.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":"superpixels-an-evaluation-of-the-state-of-the","repo_url":"https://github.com/davidstutz/superpixel-benchmark","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"superpixels-an-evaluation-of-the-state-of-the","repo_url":"https://github.com/davidstutz/cviu2018-superpixels","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"superpixels","task_name":"Superpixels"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1612.01601","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}