{"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/lsda-large-scale-detection-through-adaptation","title":"LSDA: Large Scale Detection Through Adaptation","arxiv_id":"1407.5035","date":"2014-07-18","proceeding":"NeurIPS 2014 12","authors":["Judy Hoffman","Sergio Guadarrama","Eric Tzeng","Ronghang Hu","Jeff Donahue","Ross Girshick","Trevor Darrell","Kate Saenko"],"abstract":"A major challenge in scaling object detection is the difficulty of obtaining\nlabeled images for large numbers of categories. Recently, deep convolutional\nneural networks (CNNs) have emerged as clear winners on object classification\nbenchmarks, in part due to training with 1.2M+ labeled classification images.\nUnfortunately, only a small fraction of those labels are available for the\ndetection task. It is much cheaper and easier to collect large quantities of\nimage-level labels from search engines than it is to collect detection data and\nlabel it with precise bounding boxes. In this paper, we propose Large Scale\nDetection through Adaptation (LSDA), an algorithm which learns the difference\nbetween the two tasks and transfers this knowledge to classifiers for\ncategories without bounding box annotated data, turning them into detectors.\nOur method has the potential to enable detection for the tens of thousands of\ncategories that lack bounding box annotations, yet have plenty of\nclassification data. Evaluation on the ImageNet LSVRC-2013 detection challenge\ndemonstrates the efficacy of our approach. This algorithm enables us to produce\na >7.6K detector by using available classification data from leaf nodes in the\nImageNet tree. We additionally demonstrate how to modify our architecture to\nproduce a fast detector (running at 2fps for the 7.6K detector). Models and\nsoftware are available at","url_abs":"http://arxiv.org/abs/1407.5035v3","url_pdf":"http://arxiv.org/pdf/1407.5035v3.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":"lsda-large-scale-detection-through-adaptation","repo_url":"https://github.com/jhoffman/lsda","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1407.5035","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}