{"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/object-detection-at-200-frames-per-second","title":"Object detection at 200 Frames Per Second","arxiv_id":"1805.06361","date":"2018-05-16","proceeding":null,"authors":["Rakesh Mehta","Cemalettin Ozturk"],"abstract":"In this paper, we propose an efficient and fast object detector which can\nprocess hundreds of frames per second. To achieve this goal we investigate\nthree main aspects of the object detection framework: network architecture,\nloss function and training data (labeled and unlabeled). In order to obtain\ncompact network architecture, we introduce various improvements, based on\nrecent work, to develop an architecture which is computationally light-weight\nand achieves a reasonable performance. To further improve the performance,\nwhile keeping the complexity same, we utilize distillation loss function. Using\ndistillation loss we transfer the knowledge of a more accurate teacher network\nto proposed light-weight student network. We propose various innovations to\nmake distillation efficient for the proposed one stage detector pipeline:\nobjectness scaled distillation loss, feature map non-maximal suppression and a\nsingle unified distillation loss function for detection. Finally, building upon\nthe distillation loss, we explore how much can we push the performance by\nutilizing the unlabeled data. We train our model with unlabeled data using the\nsoft labels of the teacher network. Our final network consists of 10x fewer\nparameters than the VGG based object detection network and it achieves a speed\nof more than 200 FPS and proposed changes improve the detection accuracy by 14\nmAP over the baseline on Pascal dataset.","url_abs":"http://arxiv.org/abs/1805.06361v1","url_pdf":"http://arxiv.org/pdf/1805.06361v1.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":"object-detection-at-200-frames-per-second","repo_url":"https://github.com/PaddlePaddle/PaddleDetection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}