{"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/fast-deep-stereo-with-2d-convolutional","title":"Fast Deep Stereo with 2D Convolutional Processing of Cost Signatures","arxiv_id":"1903.04939","date":"2019-03-08","proceeding":null,"authors":["Kyle Yee","Ayan Chakrabarti"],"abstract":"Modern neural network-based algorithms are able to produce highly accurate\ndepth estimates from stereo image pairs, nearly matching the reliability of\nmeasurements from more expensive depth sensors. However, this accuracy comes\nwith a higher computational cost since these methods use network architectures\ndesigned to compute and process matching scores across all candidate matches at\nall locations, with floating point computations repeated across a match volume\nwith dimensions corresponding to both space and disparity. This leads to longer\nrunning times to process each image pair, making them impractical for real-time\nuse in robots and autonomous vehicles. We propose a new stereo algorithm that\nemploys a significantly more efficient network architecture. Our method builds\nan initial match cost volume using traditional matching costs that are fast to\ncompute, and trains a network to estimate disparity from this volume.\nCrucially, our network only employs per-pixel and two-dimensional convolution\noperations: to summarize the match information at each location as a\nlow-dimensional feature vector, and to spatially process these `cost-signature'\nfeatures to produce a dense disparity map. Experimental results on the KITTI\nbenchmark show that our method delivers competitive accuracy at significantly\nhigher speeds---running at 48 frames per second on a modern GPU.","url_abs":"http://arxiv.org/abs/1903.04939v1","url_pdf":"http://arxiv.org/pdf/1903.04939v1.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":"fast-deep-stereo-with-2d-convolutional","repo_url":"https://github.com/ayanc/fdscs","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"autonomous-vehicles","task_name":"Autonomous Vehicles"},{"task_slug":null,"task_name":"GPU"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1903.04939","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}