{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,24]],"date-time":"2026-03-24T15:53:59Z","timestamp":1774367639171,"version":"3.50.1"},"reference-count":102,"publisher":"Association for Computing Machinery (ACM)","issue":"11","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Proc. VLDB Endow."],"published-print":{"date-parts":[[2017,8]]},"abstract":"<jats:p>\n            Recent advances in computer vision---in the form of deep neural networks---have made it possible to query increasing volumes of video data with high accuracy. However, neural network inference is computationally expensive at scale: applying a state-of-the-art object detector in real time (i.e., 30+ frames per second) to a single video requires a $4000 GPU. In response, we present N\n            <jats:sc>o<\/jats:sc>\n            S\n            <jats:sc>cope<\/jats:sc>\n            , a system for querying videos that can reduce the cost of neural network video analysis by up to three orders of magnitude via\n            <jats:italic>inference-optimized model search.<\/jats:italic>\n            Given a target video, object to detect, and reference neural network, N\n            <jats:sc>o<\/jats:sc>\n            S\n            <jats:sc>cope<\/jats:sc>\n            automatically searches for and trains a sequence, or cascade, of models that preserves the accuracy of the reference network but is specialized to the target video and are therefore far less computationally expensive. N\n            <jats:sc>o<\/jats:sc>\n            S\n            <jats:sc>cope<\/jats:sc>\n            cascades two types of models:\n            <jats:italic>specialized models<\/jats:italic>\n            that forego the full generality of the reference model but faithfully mimic its behavior for the target video and object; and\n            <jats:italic>difference detectors<\/jats:italic>\n            that highlight temporal differences across frames. We show that the optimal cascade architecture differs across videos and objects, so N\n            <jats:sc>o<\/jats:sc>\n            S\n            <jats:sc>cope<\/jats:sc>\n            uses an efficient cost-based optimizer to search across models and cascades. With this approach, N\n            <jats:sc>o<\/jats:sc>\n            S\n            <jats:sc>cope<\/jats:sc>\n            achieves two to three order of magnitude speed-ups (265-15,500x real-time) on binary classification tasks over fixed-angle webcam and surveillance video while maintaining accuracy within 1--5% of state-of-the-art neural networks.\n          <\/jats:p>","DOI":"10.14778\/3137628.3137664","type":"journal-article","created":{"date-parts":[[2017,9,7]],"date-time":"2017-09-07T13:35:53Z","timestamp":1504791353000},"page":"1586-1597","source":"Crossref","is-referenced-by-count":288,"title":["NoScope"],"prefix":"10.14778","volume":"10","author":[{"given":"Daniel","family":"Kang","sequence":"first","affiliation":[{"name":"Stanford InfoLab"}]},{"given":"John","family":"Emmons","sequence":"additional","affiliation":[{"name":"Stanford InfoLab"}]},{"given":"Firas","family":"Abuzaid","sequence":"additional","affiliation":[{"name":"Stanford InfoLab"}]},{"given":"Peter","family":"Bailis","sequence":"additional","affiliation":[{"name":"Stanford InfoLab"}]},{"given":"Matei","family":"Zaharia","sequence":"additional","affiliation":[{"name":"Stanford InfoLab"}]}],"member":"320","published-online":{"date-parts":[[2017,8]]},"reference":[{"key":"e_1_2_1_1_1","unstructured":"Typical cnn architecture. Creative Commons Attribution-Share Alike 4.0 International Wikimedia Commons.  Typical cnn architecture. Creative Commons Attribution-Share Alike 4.0 International Wikimedia Commons."},{"key":"e_1_2_1_2_1","unstructured":"Cisco VNI forecast and methodology 2015--2020. Technical report 2016.  Cisco VNI forecast and methodology 2015--2020. Technical report 2016."},{"key":"e_1_2_1_3_1","unstructured":"2017. https:\/\/fortunelords.com\/youtube-statistics\/.  2017. https:\/\/fortunelords.com\/youtube-statistics\/."},{"key":"e_1_2_1_4_1","volume-title":"CIDR","author":"Abadi D. J.","year":"2005"},{"key":"e_1_2_1_5_1","doi-asserted-by":"publisher","DOI":"10.1109\/TITS.2008.922938"},{"key":"e_1_2_1_6_1","doi-asserted-by":"publisher","DOI":"10.1145\/951676.951682"},{"key":"e_1_2_1_7_1","doi-asserted-by":"publisher","DOI":"10.1145\/166266.166297"},{"key":"e_1_2_1_8_1","volume-title":"NIPS","author":"Ba J.","year":"2014"},{"key":"e_1_2_1_9_1","doi-asserted-by":"publisher","DOI":"10.1007\/s00778-004-0132-6"},{"key":"e_1_2_1_10_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2010.226"},{"key":"e_1_2_1_11_1","doi-asserted-by":"publisher","DOI":"10.1145\/1007568.1007615"},{"key":"e_1_2_1_12_1","doi-asserted-by":"publisher","DOI":"10.1145\/3035918.3035928"},{"key":"e_1_2_1_13_1","volume-title":"CIDR","author":"Bailis P.","year":"2017"},{"key":"e_1_2_1_14_1","doi-asserted-by":"publisher","DOI":"10.1109\/72.165589"},{"key":"e_1_2_1_15_1","volume-title":"ECCV","author":"Benenson R.","year":"2014"},{"key":"e_1_2_1_16_1","doi-asserted-by":"publisher","DOI":"10.1006\/jvci.1997.0404"},{"key":"e_1_2_1_17_1","doi-asserted-by":"publisher","DOI":"10.1145\/1150402.1150464"},{"key":"e_1_2_1_18_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.384"},{"key":"e_1_2_1_19_1","volume-title":"CIDR","author":"Chandrasekaran S.","year":"2003"},{"key":"e_1_2_1_20_1","volume-title":"VLDB","author":"Chaudhuri S.","year":"2007"},{"key":"e_1_2_1_21_1","doi-asserted-by":"publisher","DOI":"10.1145\/320248.320249"},{"key":"e_1_2_1_22_1","volume-title":"ICML","author":"Chen W.","year":"2015"},{"key":"e_1_2_1_23_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.327"},{"key":"e_1_2_1_24_1","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2017.2669880"},{"key":"e_1_2_1_25_1","unstructured":"D. Crankshaw etal The missing piece in complex analytics: Low latency scalable model management and serving with Velox 2015.  D. Crankshaw et al. The missing piece in complex analytics: Low latency scalable model management and serving with Velox 2015."},{"key":"e_1_2_1_26_1","doi-asserted-by":"publisher","DOI":"10.1145\/872757.872838"},{"key":"e_1_2_1_27_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46454-1_29"},{"key":"e_1_2_1_28_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2011.155"},{"key":"e_1_2_1_29_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298878"},{"key":"e_1_2_1_30_1","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2016.2570556"},{"key":"e_1_2_1_31_1","doi-asserted-by":"publisher","DOI":"10.1038\/nature21056"},{"key":"e_1_2_1_32_1","unstructured":"M. Everingham etal The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results.  M. Everingham et al. The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results."},{"key":"e_1_2_1_33_1","volume-title":"NIPS","author":"Feichtenhofer C.","year":"2016"},{"key":"e_1_2_1_34_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2009.167"},{"key":"e_1_2_1_35_1","doi-asserted-by":"publisher","DOI":"10.1145\/2213836.2213874"},{"key":"e_1_2_1_36_1","doi-asserted-by":"publisher","DOI":"10.1109\/2.410146"},{"key":"e_1_2_1_37_1","doi-asserted-by":"publisher","DOI":"10.5555\/2354409.2354978"},{"key":"e_1_2_1_38_1","volume-title":"ICDE","author":"Gibbs S.","year":"1993"},{"key":"e_1_2_1_39_1","doi-asserted-by":"publisher","DOI":"10.1145\/1182807.1182886"},{"key":"e_1_2_1_40_1","volume-title":"MIT Press","author":"Goodfellow I.","year":"2016"},{"key":"e_1_2_1_41_1","volume-title":"ICLR","author":"Han S.","year":"2016"},{"key":"e_1_2_1_42_1","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2016.30"},{"key":"e_1_2_1_43_1","unstructured":"W. Han etal Seq-nms for video object detection. CoRR abs\/1602.08465 2016.  W. Han et al. Seq-nms for video object detection. CoRR abs\/1602.08465 2016."},{"key":"e_1_2_1_44_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_2_1_45_1","volume-title":"NIPS","author":"Hinton G.","year":"2014"},{"key":"e_1_2_1_46_1","unstructured":"G. Hinton and T. Tieleman. Lecture 6.5 - rmsprop. Technical report 2012.  G. Hinton and T. Tieleman. Lecture 6.5 - rmsprop. Technical report 2012."},{"key":"e_1_2_1_47_1","volume-title":"CIDR","author":"Idreos S.","year":"2007"},{"key":"e_1_2_1_48_1","doi-asserted-by":"publisher","DOI":"10.1145\/190627.190638"},{"key":"e_1_2_1_49_1","doi-asserted-by":"publisher","DOI":"10.1007\/s005300050087"},{"key":"e_1_2_1_50_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.95"},{"key":"e_1_2_1_51_1","unstructured":"K. Kang etal T-cnn: Tubelets with convolutional neural networks for object detection from videos. arXiv preprint arXiv:1604.02532 2016.  K. Kang et al. T-cnn: Tubelets with convolutional neural networks for object detection from videos. arXiv preprint arXiv:1604.02532 2016."},{"key":"e_1_2_1_52_1","doi-asserted-by":"crossref","unstructured":"V. Kastrinaki M. Zervakis and K. Kalaitzakis. A survey of video processing techniques for traffic applications. Image and vision computing 2003.  V. Kastrinaki M. Zervakis and K. Kalaitzakis. A survey of video processing techniques for traffic applications. Image and vision computing 2003.","DOI":"10.1016\/S0262-8856(03)00004-0"},{"key":"e_1_2_1_53_1","doi-asserted-by":"publisher","DOI":"10.1016\/S0167-8655(02)00194-0"},{"key":"e_1_2_1_54_1","volume-title":"CIDR","author":"Kraska T.","year":"2013"},{"key":"e_1_2_1_55_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-48881-3_54"},{"key":"e_1_2_1_56_1","volume-title":"NIPS","author":"Krizhevsky A.","year":"2012"},{"key":"e_1_2_1_57_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.1996.543585"},{"key":"e_1_2_1_58_1","doi-asserted-by":"publisher","DOI":"10.1038\/nature14539"},{"key":"e_1_2_1_59_1","doi-asserted-by":"publisher","DOI":"10.1145\/1066157.1066239"},{"key":"e_1_2_1_60_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7299170"},{"key":"e_1_2_1_61_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW.2015.7301269"},{"key":"e_1_2_1_62_1","doi-asserted-by":"crossref","unstructured":"T.-Y. Lin etal Microsoft coco: Common objects in context. September 2014.  T.-Y. Lin et al. Microsoft coco: Common objects in context. September 2014.","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"e_1_2_1_63_1","doi-asserted-by":"publisher","DOI":"10.5555\/850924.851523"},{"key":"e_1_2_1_64_1","doi-asserted-by":"publisher","DOI":"10.1080\/01431160600746456"},{"key":"e_1_2_1_65_1","doi-asserted-by":"publisher","DOI":"10.1186\/s13640-016-0143-4"},{"key":"e_1_2_1_66_1","doi-asserted-by":"publisher","DOI":"10.1145\/3065384"},{"key":"e_1_2_1_67_1","volume-title":"Wired","author":"Metz C.","year":"2017"},{"key":"e_1_2_1_68_1","doi-asserted-by":"publisher","DOI":"10.1109\/DICTA.2016.7797042"},{"key":"e_1_2_1_69_1","doi-asserted-by":"publisher","DOI":"10.1145\/1265530.1265561"},{"key":"e_1_2_1_70_1","doi-asserted-by":"publisher","DOI":"10.1109\/34.291452"},{"key":"e_1_2_1_71_1","unstructured":"H. Nam etal Modeling and propagating cnns in a tree structure for visual tracking. arXiv:1608.07242 2016.  H. Nam et al. Modeling and propagating cnns in a tree structure for visual tracking. arXiv:1608.07242 2016."},{"key":"e_1_2_1_72_1","doi-asserted-by":"publisher","DOI":"10.1109\/2.410150"},{"key":"e_1_2_1_73_1","doi-asserted-by":"publisher","DOI":"10.1145\/335191.335436"},{"key":"e_1_2_1_74_1","unstructured":"S. A. Papert. The summer vision project. 1966.  S. A. Papert. The summer vision project. 1966."},{"key":"e_1_2_1_75_1","volume-title":"2016 GPU Technology Conference (https:\/\/goo.gl\/Bdjdgm)","author":"Patait A.","year":"2016"},{"key":"e_1_2_1_76_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2007.383172"},{"key":"e_1_2_1_77_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_32"},{"key":"e_1_2_1_78_1","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2016.32"},{"key":"e_1_2_1_79_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.91"},{"key":"e_1_2_1_80_1","doi-asserted-by":"crossref","unstructured":"J. Redmon and A. Farhadi. Yolo9000: Better faster stronger. arXiv preprint arXiv:1612.08242 2016.  J. Redmon and A. Farhadi. Yolo9000: Better faster stronger. arXiv preprint arXiv:1612.08242 2016.","DOI":"10.1109\/CVPR.2017.690"},{"key":"e_1_2_1_81_1","volume-title":"NIPS","author":"Ren S.","year":"2015"},{"key":"e_1_2_1_82_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"e_1_2_1_83_1","doi-asserted-by":"publisher","DOI":"10.1145\/582095.582099"},{"key":"e_1_2_1_84_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2013.465"},{"key":"e_1_2_1_85_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2013.446"},{"key":"e_1_2_1_86_1","volume-title":"Springer Science & Business Media","author":"Szeliski R.","year":"2010"},{"key":"e_1_2_1_87_1","doi-asserted-by":"publisher","DOI":"10.1109\/ITSC.2011.6083125"},{"key":"e_1_2_1_88_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-013-0620-5"},{"key":"e_1_2_1_89_1","doi-asserted-by":"publisher","DOI":"10.1007\/s00138-007-0084-0"},{"key":"e_1_2_1_90_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2001.990517"},{"key":"e_1_2_1_91_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.362"},{"key":"e_1_2_1_92_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2013.312"},{"key":"e_1_2_1_93_1","doi-asserted-by":"publisher","DOI":"10.1145\/1177352.1177355"},{"key":"e_1_2_1_94_1","doi-asserted-by":"publisher","DOI":"10.1109\/69.755617"},{"key":"e_1_2_1_95_1","doi-asserted-by":"publisher","DOI":"10.1038\/ncomms12474"},{"key":"e_1_2_1_96_1","first-page":"53","volume-title":"CVPR","author":"Yue-Hei Ng J.","year":"2015"},{"key":"e_1_2_1_97_1","volume-title":"NSDI","author":"Zhang H.","year":"2017"},{"key":"e_1_2_1_98_1","volume-title":"CVPR","author":"Zhanget S.","year":"2016"},{"key":"e_1_2_1_99_1","doi-asserted-by":"publisher","DOI":"10.1145\/954339.954342"},{"key":"e_1_2_1_100_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2006.119"},{"key":"e_1_2_1_101_1","doi-asserted-by":"publisher","DOI":"10.1109\/TKDE.2005.83"},{"key":"e_1_2_1_102_1","volume-title":"ICLR","author":"Zoph B.","year":"2017"}],"container-title":["Proceedings of the VLDB Endowment"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.14778\/3137628.3137664","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,12,28]],"date-time":"2022-12-28T10:00:04Z","timestamp":1672221604000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.14778\/3137628.3137664"}},"subtitle":["optimizing neural network queries over video at scale"],"short-title":[],"issued":{"date-parts":[[2017,8]]},"references-count":102,"journal-issue":{"issue":"11","published-print":{"date-parts":[[2017,8]]}},"alternative-id":["10.14778\/3137628.3137664"],"URL":"https:\/\/doi.org\/10.14778\/3137628.3137664","relation":{},"ISSN":["2150-8097"],"issn-type":[{"value":"2150-8097","type":"print"}],"subject":[],"published":{"date-parts":[[2017,8]]}}}