计算机科学 ›› 2023, Vol. 50 ›› Issue (6A): 220500230-9.doi: 10.11896/jsjkx.220500230
王国刚, 吴艳, 刘一博
WANG Guogang, WU Yan, LIU Yibo
摘要: 针对EfficientDet算法鲁棒性低、回归损失函数收敛性能差、标签边缘化问题,提出了回归收敛缩放混合的深度迭代复合缩放CNN目标检测算法。该算法采用2×2缩放混合正则化方法,增强训练样本,避免训练过拟合,提高模型泛化能力;利用完全交并比损失抑制冗余预测框,将中心点距离和纵横比作为边界框坐标预测的损失函数惩罚项,使卷积神经网络回归更准确,提高了收敛速度和定位精度;设置平滑参数,对边缘化标签分布和均匀分布加权求和生成标签平滑正则化分布,建立类标签平滑交叉熵损失,提高模型的标签容错率。实验结果表明,所提算法的均值平均精度为88.31%,网络模型参数个数为8.10×106,相比EfficientDet-D2算法,均值平均精度提高了3.29%,网络模型参数个数没有增加,相比YOLOv4,YOLOv3,SSD,Faster R-CNN和Fast R-CNN算法,均值平均精度分别提升了5.2%,10.71 %,14.01%,15.11% 和18.30 %,网络模型参数个数分别减少了55.94×106,52.91×106,16.09×106,55.18×106和53.11×106。所提目标检测模型,提高了检测准确度和F1得分;检测每张测试图片仅需0.73s,满足实时性要求。
中图分类号:
[1]ARRINGTON M,ELBICH D,DAI J,et al.Introducing the female Cambridge face memory test-long form(F-CFMT+)[J].Behavior Research Methods,2022:1-14. [2]YU J,HAO X,CUI Z,et al.Boosting Fairness for Masked Face Recognition[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2021:1531-1540. [3]SUNG C S,PARK J Y.Design of an intelligent video surveil-lance system for crime prevention:applying deep learning technology[J].Multimedia Tools and Applications,2021,80(26):34297-34309. [4]KIM J S,KIM M G,PAN S B.A study on implementation ofreal-time intelligent video surveillance system based on embedded module[J].EURASIP Journal on Image and Video Proces-sing,2021,2021(1):1-22. [5]XIAOFENG T.Ecological driving on multiphase trajectories and multiobjective optimization for autonomous electric vehicle platoon[J].Scientific Reports,2022,12(1):1-16. [6]TIAN X,LIU J,MALLICK M,et al.Simultaneous detection and tracking of moving-target shadows in ViSAR imagery[J].IEEE Transactions on Geoscience and Remote Sensing,2020,59(2):1182-1199. [7]LIU S,WANG S,LIU X,et al.Fuzzy detection aided real-time and robust visual tracking under complex environments[J].IEEE Transactions on Fuzzy Systems,2020,29(1):90-102. [8]FELZENSZWALB P F,GIRSHICK R B,MCALLESTER D,et al.Object detection with discriminatively trained part-based models[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2009,32(9):1627-1645. [9]FRANCOIS C.Xception:Deep learning with depth wise separable convolutions[C]//CVPR.2017:1800-1807. [10]REDMON J,DIVVALA S,GIRSHICK R,et al.You only look once:Unified,real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition.New York,USA:IEEE,2016.779-788. [11]REDMON J,FARHADI A.YOLO9000:better,faster,stronger[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Honolulu,USA:IEEE,2017.6517-6525. [12]REDMON JOSEPH,FARHADI A.Yolov3:An incremental improvement[J].arXiv:1804.02767,2018. [13]FU C Y,LIU W,RANGA A,et al.DSSD:Deconvolutional single shot detector[J].arXiv:1701.06659,2017. [14]SHEN Z Q,LIU Z H,LI J G,et al.DSOD:Learning deeply supervised object detectors from scratch[C]//Proceedings of the IEEE International Conference on Computer Vision.New York,USA:IEEE,2017:1919-1927. [15]GIRSHICK R,DONAHUE J,DARRELL T.Rich feature hie-rarchies for accurate object detection and semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Columbus,USA:IEEE,2014:580-587. [16]REN S,HE K M,GIRSHICK R,et al.Faster R-CNN:Towards real-time object detection with region proposal networks[C]//International Conference on Neural Information Processing Systems.Cambridge USA:MIT Press,2015:91-99. [17]MINGXING TAN,RUOMING PANG,QUOC V L E.EfficientDet:Scalable and efficient object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.CVPR,2020:10778-10787. [18]SANGDOO Y,DONGYOON H,SEONG J O,et al.Cut Mix:Regularization strategy to train strong classififiers with localizable features[C]//Proceedings of the IEEE International Confe-rence on Computer Vision.ICCV,2019:6023-6032. [19]ZHENG Z H,WANG P,LIU W,et al.Distance-IoU Loss:Faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artifificial Intelligence.AAAI,2020:12993-13000. [20]CHRISTIAN S,VINCENT V,SERGEY I,et al.Rethinking the inception architecture for computer vision[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.CVPR,2016:2818-2826. [21]TAN M,LE Q.Efficientnet:Rethinking model scaling for convolutional neural networks[C]//International Conference on Machine Learning.PMLR,2019:6105-6114. |
|