{"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/pvanet-deep-but-lightweight-neural-networks","title":"PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection","arxiv_id":"1608.08021","date":"2016-08-29","proceeding":null,"authors":["Kye-Hyeon Kim","Sanghoon Hong","Byungseok Roh","Yeongjae Cheon","Minje Park"],"abstract":"This paper presents how we can achieve the state-of-the-art accuracy in\nmulti-category object detection task while minimizing the computational cost by\nadapting and combining recent technical innovations. Following the common\npipeline of \"CNN feature extraction + region proposal + RoI classification\", we\nmainly redesign the feature extraction part, since region proposal part is not\ncomputationally expensive and classification part can be efficiently compressed\nwith common techniques like truncated SVD. Our design principle is \"less\nchannels with more layers\" and adoption of some building blocks including\nconcatenated ReLU, Inception, and HyperNet. The designed network is deep and\nthin and trained with the help of batch normalization, residual connections,\nand learning rate scheduling based on plateau detection. We obtained solid\nresults on well-known object detection benchmarks: 83.8% mAP (mean average\nprecision) on VOC2007 and 82.5% mAP on VOC2012 (2nd place), while taking only\n750ms/image on Intel i7-6700K CPU with a single core and 46ms/image on NVIDIA\nTitan X GPU. Theoretically, our network requires only 12.3% of the\ncomputational cost compared to ResNet-101, the winner on VOC2012.","url_abs":"http://arxiv.org/abs/1608.08021v3","url_pdf":"http://arxiv.org/pdf/1608.08021v3.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":"pvanet-deep-but-lightweight-neural-networks","repo_url":"https://github.com/sanghoon/pva-faster-rcnn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"caffe2","reach":null},{"paper_slug":"pvanet-deep-but-lightweight-neural-networks","repo_url":"https://github.com/tonyhsu32/AI-Cafe-with-machine-learning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":null,"task_name":"CPU"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"real-time-object-detection","task_name":"Real-Time Object Detection"},{"task_slug":"region-proposal","task_name":"Region Proposal"},{"task_slug":"scheduling","task_name":"Scheduling"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"relu","method_name":"ReLU"}],"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}