{"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/reconfiguring-the-imaging-pipeline-for","title":"Reconfiguring the Imaging Pipeline for Computer Vision","arxiv_id":"1705.04352","date":"2017-05-11","proceeding":"ICCV 2017 10","authors":["Mark Buckler","Suren Jayasuriya","Adrian Sampson"],"abstract":"Advancements in deep learning have ignited an explosion of research on\nefficient hardware for embedded computer vision. Hardware vision acceleration,\nhowever, does not address the cost of capturing and processing the image data\nthat feeds these algorithms. We examine the role of the image signal processing\n(ISP) pipeline in computer vision to identify opportunities to reduce\ncomputation and save energy. The key insight is that imaging pipelines should\nbe designed to be configurable: to switch between a traditional photography\nmode and a low-power vision mode that produces lower-quality image data\nsuitable only for computer vision. We use eight computer vision algorithms and\na reversible pipeline simulation tool to study the imaging system's impact on\nvision performance. For both CNN-based and classical vision algorithms, we\nobserve that only two ISP stages, demosaicing and gamma compression, are\ncritical for task performance. We propose a new image sensor design that can\ncompensate for skipping these stages. The sensor design features an adjustable\nresolution and tunable analog-to-digital converters (ADCs). Our proposed\nimaging system's vision mode disables the ISP entirely and configures the\nsensor to produce subsampled, lower-precision image data. This vision mode can\nsave ~75% of the average energy of a baseline photography mode while having\nonly a small impact on vision task accuracy.","url_abs":"http://arxiv.org/abs/1705.04352v3","url_pdf":"http://arxiv.org/pdf/1705.04352v3.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":"reconfiguring-the-imaging-pipeline-for","repo_url":"https://github.com/cucapra/approx-vision","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"demosaicking","task_name":"Demosaicking"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.04352","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}