{"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/on-the-performance-of-convnet-features-for","title":"On the Performance of ConvNet Features for Place Recognition","arxiv_id":"1501.04158","date":"2015-01-17","proceeding":null,"authors":["Niko Sünderhauf","Feras Dayoub","Sareh Shirazi","Ben Upcroft","Michael Milford"],"abstract":"After the incredible success of deep learning in the computer vision domain,\nthere has been much interest in applying Convolutional Network (ConvNet)\nfeatures in robotic fields such as visual navigation and SLAM. Unfortunately,\nthere are fundamental differences and challenges involved. Computer vision\ndatasets are very different in character to robotic camera data, real-time\nperformance is essential, and performance priorities can be different. This\npaper comprehensively evaluates and compares the utility of three\nstate-of-the-art ConvNets on the problems of particular relevance to navigation\nfor robots; viewpoint-invariance and condition-invariance, and for the first\ntime enables real-time place recognition performance using ConvNets with large\nmaps by integrating a variety of existing (locality-sensitive hashing) and\nnovel (semantic search space partitioning) optimization techniques. We present\nextensive experiments on four real world datasets cultivated to evaluate each\nof the specific challenges in place recognition. The results demonstrate that\nspeed-ups of two orders of magnitude can be achieved with minimal accuracy\ndegradation, enabling real-time performance. We confirm that networks trained\nfor semantic place categorization also perform better at (specific) place\nrecognition when faced with severe appearance changes and provide a reference\nfor which networks and layers are optimal for different aspects of the place\nrecognition problem.","url_abs":"http://arxiv.org/abs/1501.04158v3","url_pdf":"http://arxiv.org/pdf/1501.04158v3.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":"on-the-performance-of-convnet-features-for","repo_url":"https://github.com/aghagol/loop-detection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"visual-navigation","task_name":"Visual Navigation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1501.04158","atlas_url":"https://app.syntology.ai/?focus=1501.04158","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}