第3期 葛园园，等：自动驾驶场景下小且密集的交通标志检测 ·371· (a)conv5 3 到 (b)conv4 3 (c)conv41、conv42、conv43 图5使用不同层及组合层检测结果 Fig.5 Using different layers and aggregated-layer-detection results 参考文献： 2826. [8]SAHA S,SINGH G,SAPIENZA M,et al.Deep learning for [1]Krizhevsky A,Sutskever I,Hinton G E.Imagenet classifica- detecting multiple space-time action tubes in videos[J].arX- tion with deep convolutional neural networks[Cl//Proceed- iv:1608.01529,2016. ings of Advances in Neural Information Processing Sys- [9]GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich fea- tems.Stateline,NV,USA,2012:1097-1105. ture hierarchies for accurate object detection and semantic [2]SZEGEDY C,LIU Wei,JIA Yangqing,et al.Going deeper segmentation[C]//Proceedings of 2014 IEEE Conference on with convolutions[C]//Proceedings of 2015 IEEE Confer- Computer Vision and Pattern Recognition.Columbus,OH, ence on Computer Vision and Pattern Recognition.Boston, USA.2014:580-587 MA,USA,2015:1-9, [3]CHEN L C,PAPANDREOU G,KOKKINOS I,et al.Se- [10]UIJLINGS J RR,VAN DE SANDE K E A,GEVERS T, et al.Selective search for object recognition[J].Internation- mantic image segmentation with deep convolutional nets and fully connected CRFs[J].arXiv:1412.7062,2015. al journal of computer vision,2013,104(2):154-171 [4]YU Gang,YUAN Junsong.Fast action proposals for hu- [11]EVERINGHAM M,VAN GOOL L,WILLIAMS C K I,et man action detection and search[C]//Proceedings of 2015 al.The Pascal visual object classes(VOC)challenge[J].In- IEEE Conference on Computer Vision and Pattern Recogni- ternational journal of computer vision,2010,88(2): tion.Boston,MA,USA,2015:1302-1311. 303-338 [5]HONG S,YOU T,KWAK S,et al.Online tracking by [12]HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al.Spa- learning discriminative saliency map with convolutional tial pyramid pooling in deep convolutional networks for neural network[Cl//Proceedings of the 32nd International visual recognition[C]//Proceedings of the 13th European Conference on Machine Learning.Lille,France,2015: Conference on Computer Vision.Zurich,Switzerland, 597-606. 2014:346-361. [6]WANG Naiyan,YEUNG D Y.Learning a deep compact [13]GIRSHICK R.Fast R-CNN[C]//Proceedings of 2015 IEEE image representation for visual tracking[Cl//Proceedings of International Conference on Computer Vision.Santiago, the 26th International Conference on Neural Information Chile,.2015:1440-1448. Processing Systems.Lake Tahoe,USA,2013:809-817. [14]Ren S.He K.Girshick R,et al.Faster r-cnn:Towards real- [7]SZEGEDY C,VANHOUCKE V,IOFFE S,et al.Rethink- time object detection with region proposal networks[C]// ing the inception architecture for computer vision[C]//Pro- Proceedings of 2015 Advances in Neural Information Pro- ceedings of 2016 IEEE Conference on Computer Vision and cessing Systems.Montreal,Canada,2015:91-99 Pattern Recognition.Las Vegas,NV,USA,2016:2818- [15]Li Y,He K,Sun J.R-fcn:Object detection via region-参考文献： Krizhevsky A, Sutskever I, Hinton G E. Imagenet classifica￾tion with deep convolutional neural networks[C]//Proceed￾ings of Advances in Neural Information Processing Sys￾tems. Stateline, NV, USA, 2012: 1097-1105. [1] SZEGEDY C, LIU Wei, JIA Yangqing, et al. Going deeper with convolutions[C]//Proceedings of 2015 IEEE Confer￾ence on Computer Vision and Pattern Recognition. Boston, MA, USA, 2015: 1–9. [2] CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Se￾mantic image segmentation with deep convolutional nets and fully connected CRFs[J]. arXiv:1412.7062, 2015. [3] YU Gang, YUAN Junsong. Fast action proposals for hu￾man action detection and search[C]//Proceedings of 2015 IEEE Conference on Computer Vision and Pattern Recogni￾tion. Boston, MA, USA, 2015: 1302–1311. [4] HONG S, YOU T, KWAK S, et al. Online tracking by learning discriminative saliency map with convolutional neural network[C]//Proceedings of the 32nd International Conference on Machine Learning. Lille, France, 2015: 597–606. [5] WANG Naiyan, YEUNG D Y. Learning a deep compact image representation for visual tracking[C]//Proceedings of the 26th International Conference on Neural Information Processing Systems. Lake Tahoe, USA, 2013: 809–817. [6] SZEGEDY C, VANHOUCKE V, IOFFE S, et al. Rethink￾ing the inception architecture for computer vision[C]//Pro￾ceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA, 2016: 2818– [7] 2826. SAHA S, SINGH G, SAPIENZA M, et al. Deep learning for detecting multiple space-time action tubes in videos[J]. arX￾iv: 1608.01529, 2016. [8] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich fea￾ture hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus, OH, USA, 2014: 580–587. [9] UIJLINGS J R R, VAN DE SANDE K E A, GEVERS T, et al. Selective search for object recognition[J]. Internation￾al journal of computer vision, 2013, 104(2): 154–171. [10] EVERINGHAM M, VAN GOOL L, WILLIAMS C K I, et al. The Pascal visual object classes (VOC) challenge[J]. In￾ternational journal of computer vision, 2010, 88(2): 303–338. [11] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Spa￾tial pyramid pooling in deep convolutional networks for visual recognition[C]//Proceedings of the 13th European Conference on Computer Vision. Zurich, Switzerland, 2014: 346–361. [12] GIRSHICK R. Fast R-CNN[C]//Proceedings of 2015 IEEE International Conference on Computer Vision. Santiago, Chile, 2015: 1440–1448. [13] Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real￾time object detection with region proposal networks[C]// Proceedings of 2015 Advances in Neural Information Pro￾cessing Systems. Montréal,Canada, 2015: 91-99. [14] [15] Li Y, He K, Sun J. R-fcn: Object detection via region- (a) conv5_3 (b) conv4_3 (c) conv4_1ȟDPOW4_2ȟDPOW4_3 图 5 使用不同层及组合层检测结果 Fig. 5 Using different layers and aggregated-layer-detection results 第 3 期 葛园园，等：自动驾驶场景下小且密集的交通标志检测 ·371·