基于CNN的改进行人重识别技术
|
摘要
针对行人重识别研究中训练样本的不足,为提高识别精度及泛化能力,提出一种基于卷积神经网络的改进行人重识别方法。首先对训练数据集进行扩充,使用生成对抗网络无监督学习方法生成无标签图像;然后与原数据集联合作半监督卷积神经网络训练,通过构建一个Siamese网络,结合分类模型和验证模型的特点进行训练;最后加入无标签图像类别分布方法,计算交叉熵损失来进行相似度量。实验结果表明,在Market-1501、CUHK03和DukeMTMC-reID数据集上,该方法相比原有的Siamese方法在Rank-1和mAP等性能指标上有近3～5个百分点的提升。当样本较少时,该方法具有一定应用价值。
For the lack of training samples in pedestrian re-identification(re-ID)research,we propose a pedestrian re-ID method based on convolutional neural network(CNN)to improve the recognition accuracy and generalization ability.Firstly,we employ the unsupervised learning method for the generative adversarial network to generate unlabeled images,so the training data set is expanded.Secondly,the original data set is collaborated to perform semi-supervised CNN training,and a Siamese network is constructed to perform training according to the features of the identification model and the verification model.Finally,the unlabeled image category distribution method is introduced,and the cross entropy loss is calculated to perform similarity measurement.Experiments on the Market-1501,CUHK03,and DukeMTMC-reID datasets show that the proposed method has a nearly 3%to 5%improvement in performance indicators such as rank-1 and mAPin comparison with the original Siamese method.The proposed method has certain application value in small sample scenarios.
For the lack of training samples in pedestrian re-identification(re-ID)research,we propose a pedestrian re-ID method based on convolutional neural network(CNN)to improve the recognition accuracy and generalization ability.Firstly,we employ the unsupervised learning method for the generative adversarial network to generate unlabeled images,so the training data set is expanded.Secondly,the original data set is collaborated to perform semi-supervised CNN training,and a Siamese network is constructed to perform training according to the features of the identification model and the verification model.Finally,the unlabeled image category distribution method is introduced,and the cross entropy loss is calculated to perform similarity measurement.Experiments on the Market-1501,CUHK03,and DukeMTMC-reID datasets show that the proposed method has a nearly 3%to 5%improvement in performance indicators such as rank-1 and mAPin comparison with the original Siamese method.The proposed method has certain application value in small sample scenarios.
引文
[1]Yang Yang,Yang Ji-mei,Yan Jun-jie,et al.Salient color names for person re-identification[C]∥Proc of the European Conference on Computer Vision(ECCV),2014:536-551.
[2]Liao Sheng-cai,Hu Yang,Zhu Xiang-yu,et al.Person re-identification by local maximal occurrence representation and metric learning[C]∥Proc of the 2015IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2015:2197-2206.
[3]Matsukawa T,Okabe T,Suzuki E,et al.Hierarchical Gaussian descriptor for person re-identification[C]∥Proc of the2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:1363-1372.
[4]Xiong Fei,Gou Meng-ran,Camps O,et al.Person re-identification using kernel-based metric learning methods[C]∥Proc of the European Conference on Computer Vision(ECCV),2014:1-16.
[5]Lisanti G,Masi I,Bagdanov A D,et al.Person re-identification by iterative re-weighted sparse ranking[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(8):1629-1642.
[6]Zheng Wei-shi,Gong Shao-gang,Xiang Tao.Reidentification by relative distance comparison[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(3):653-668.
[7]Peng Zhi-yong,Chang Fa-liang,Liu Hong-bin,et al.Person re-identification algorithm based on HSV model and keypoints matching[J].Journal of Optoelectronics·Laser,2015,26(8):1575-1582.(in Chinese)
[8]Huang Xin-yu,Xu Jiao-long,Guo Gang,et al.Real-time pedestrain reidentificaton based on enhanced aggregated channel features[J].Laser&Optoelectronics Progress,2017,54(9):113-121.(in Chinese)
[9]Sun Jin-yu,Wang Hong-yuan,Zhang Ji,et al.Person re-identification method based on block sparse representation[J].Journal of Computer Applications,2018,38(2):448-453.(in Chinese)
[10]Zagoruyko S,Komodakis N.Learning to compare image patches via convolutional neural networks[C]∥Proc of the2015IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2015:4353-4361.
[11]Yi Dong,Lei Zhen,Liao Sheng-cai,et al.Deep metric learning for person re-identification[C]∥Proc of the 2014 22nd International Conference on Pattern Recognition,2014:34-39.
[12]Varior R R,Shuai Bing,Lu Ji-wen,et al.A Siamese long short-term memory architecture for human re-identification[C]∥Proc of the European Conference on Computer Vision(ECCV),2016:135-153.
[13]Varior R R,Haloi M,Wang Gang.Gated Siamese convolutional neural network architecture for human re-identification[C]∥Proc of the European Conference on Computer Vision(ECCV),2016:791-808.
[14]Rasmus A,Valpola H,Honkala M,et al.Semi-supervised learning with ladder networks[C]∥Proc of the 28th International Conference on Neural Information Processing Systems,2015:3546-3554.
[15]Chen Hao-ren,Wang Yao-mei,Shi Ye-min,et al.Deep transfer learning for person re-identification[C]∥Proc of2018IEEE 4th International Conference on Multimedia Big Date(BigMM),2018:1-5.
[16]Zhang Shao-ting,Yang Ming,Cour T,et al.Query specific rank fusion for image retrieval[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(4):803-815.
[17]Bai Song,Bai Xiang,Tian Qi,et al.Regularized diffusion process for visual retrieval[C]∥Proc of the 31st AAAIConference on Artificial Intelligence,2017:3967-3973.
[18]Zheng Zhe-dong,Zheng Liang,Yang Yi.Unlabeled samples generated by GAN improve the person re-identification baseline in vitro[C]∥Proc of the 2017IEEE International Conference on Computer Vision(ICCV),2017:3774-3782.
[19]Goodfellow I J,Pouget-Abadie J,Mirza M,et al.Generative adversarial nets[C]∥Proc of the 27th International Conference on Neural Information Processing Systems-Volume 2,2014:2672-2680.
[20]Zheng Zhe-dong,Zheng Liang,Yang Yi.A discriminatively learned CNN embedding for person re-identification[J].ACM Transactions on Multimedia Computing Communications&Applications,2016,14(1):Article No.13.
[21]Radford A,Metz L,Chintala S.Unsupervised representation learning with deep convolutional generative adversarial networks[J].arXiv preprint arXiv:1511.06434v2,2016.
[22]Sun Yi,Chen Yu-heng,Wang Xiao-gang,et al.Deep learning face representation by joint identification-verification[C]∥Proc of the 27th International Conference on Neural Information Processing Systems,2014:1988-1996.
[23]Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:Asimple way to prevent neural networks from overfitting[J].The Journal Machine Learning Research,2014,15(1):1929-1958.
[24]He Kai-ming,Zhang Xiang-yu,Ren Shao-qing,et al.Deep residual learning for image recognition[C]∥Proc of the2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:770-778.
[25]Szegedy C,Vanhoucke V,Ioffe S,et al.Rethinking the inception architecture for computer vision[C]∥Proc of the2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:2818-2826.
[26]Zheng Liang,Shen Li-yue,Tian Lu,et al.Scalable person reidentification:A benchmark[C]∥Proc of the 2015IEEEInternational Conference on Computer Vision(ICCV),2015:1116-1124.
[27]Li Wei,Zhao Rui,Xiao Tong,et al.DeepReID:Deep filter pairing neural network for person re-identification[C]∥Proc of the 2014IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2014:152-159.
[28]Ristani E,Solera F,Zou R S,et al.Performance measures and a data set for multi-target,multi-camera tracking[C]∥Proc of the European Conference on Computer Vision(EC-CV),2016:17-35.
[29]Ustinova E,Ganin Y,Lempitsky V S.Multi-region bilinear convolutional neural networks for person re-identification[C]∥Proc of the 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance(AVSS),2017:1-6.
[30]Zhang Li,Xiang Tao,Gong Shao-gang.Learning a discriminative null space for person re-identification[C]∥Proc of the 2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:1239-1248.
[31]Barbosa I B,Cristani M,Caputo B,et al.Looking beyond appearances:Synthetic training data for deep CNNs in re-identification[J].Computer Vision and Image Understanding,2018,167:50-62.
[32]Lin Yu-tian,Zheng Liang,Zheng Zhe-dong,et al.Improving person re-identification by attribute and identity learning[J].arXiv preprint arXiv:1703.07220v2,2017.
[7]彭志勇,常发亮,刘洪彬,等.基于HSV模型和特征点匹配的行人重识别算法[J].光电子·激光,2015,26(8):1575-1582.
[8]黄新宇,许娇龙,郭纲,等.基于增强聚合通道特征的实时行人重识别[J].激光与光电子学进展,2017,54(9):113-121.
[9]孙金玉,王洪元,张继,等.基于块稀疏表示的行人重识别方法[J].计算机应用,2018,38(2):448-453.
[2]Liao Sheng-cai,Hu Yang,Zhu Xiang-yu,et al.Person re-identification by local maximal occurrence representation and metric learning[C]∥Proc of the 2015IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2015:2197-2206.
[3]Matsukawa T,Okabe T,Suzuki E,et al.Hierarchical Gaussian descriptor for person re-identification[C]∥Proc of the2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:1363-1372.
[4]Xiong Fei,Gou Meng-ran,Camps O,et al.Person re-identification using kernel-based metric learning methods[C]∥Proc of the European Conference on Computer Vision(ECCV),2014:1-16.
[5]Lisanti G,Masi I,Bagdanov A D,et al.Person re-identification by iterative re-weighted sparse ranking[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(8):1629-1642.
[6]Zheng Wei-shi,Gong Shao-gang,Xiang Tao.Reidentification by relative distance comparison[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(3):653-668.
[7]Peng Zhi-yong,Chang Fa-liang,Liu Hong-bin,et al.Person re-identification algorithm based on HSV model and keypoints matching[J].Journal of Optoelectronics·Laser,2015,26(8):1575-1582.(in Chinese)
[8]Huang Xin-yu,Xu Jiao-long,Guo Gang,et al.Real-time pedestrain reidentificaton based on enhanced aggregated channel features[J].Laser&Optoelectronics Progress,2017,54(9):113-121.(in Chinese)
[9]Sun Jin-yu,Wang Hong-yuan,Zhang Ji,et al.Person re-identification method based on block sparse representation[J].Journal of Computer Applications,2018,38(2):448-453.(in Chinese)
[10]Zagoruyko S,Komodakis N.Learning to compare image patches via convolutional neural networks[C]∥Proc of the2015IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2015:4353-4361.
[11]Yi Dong,Lei Zhen,Liao Sheng-cai,et al.Deep metric learning for person re-identification[C]∥Proc of the 2014 22nd International Conference on Pattern Recognition,2014:34-39.
[12]Varior R R,Shuai Bing,Lu Ji-wen,et al.A Siamese long short-term memory architecture for human re-identification[C]∥Proc of the European Conference on Computer Vision(ECCV),2016:135-153.
[13]Varior R R,Haloi M,Wang Gang.Gated Siamese convolutional neural network architecture for human re-identification[C]∥Proc of the European Conference on Computer Vision(ECCV),2016:791-808.
[14]Rasmus A,Valpola H,Honkala M,et al.Semi-supervised learning with ladder networks[C]∥Proc of the 28th International Conference on Neural Information Processing Systems,2015:3546-3554.
[15]Chen Hao-ren,Wang Yao-mei,Shi Ye-min,et al.Deep transfer learning for person re-identification[C]∥Proc of2018IEEE 4th International Conference on Multimedia Big Date(BigMM),2018:1-5.
[16]Zhang Shao-ting,Yang Ming,Cour T,et al.Query specific rank fusion for image retrieval[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(4):803-815.
[17]Bai Song,Bai Xiang,Tian Qi,et al.Regularized diffusion process for visual retrieval[C]∥Proc of the 31st AAAIConference on Artificial Intelligence,2017:3967-3973.
[18]Zheng Zhe-dong,Zheng Liang,Yang Yi.Unlabeled samples generated by GAN improve the person re-identification baseline in vitro[C]∥Proc of the 2017IEEE International Conference on Computer Vision(ICCV),2017:3774-3782.
[19]Goodfellow I J,Pouget-Abadie J,Mirza M,et al.Generative adversarial nets[C]∥Proc of the 27th International Conference on Neural Information Processing Systems-Volume 2,2014:2672-2680.
[20]Zheng Zhe-dong,Zheng Liang,Yang Yi.A discriminatively learned CNN embedding for person re-identification[J].ACM Transactions on Multimedia Computing Communications&Applications,2016,14(1):Article No.13.
[21]Radford A,Metz L,Chintala S.Unsupervised representation learning with deep convolutional generative adversarial networks[J].arXiv preprint arXiv:1511.06434v2,2016.
[22]Sun Yi,Chen Yu-heng,Wang Xiao-gang,et al.Deep learning face representation by joint identification-verification[C]∥Proc of the 27th International Conference on Neural Information Processing Systems,2014:1988-1996.
[23]Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:Asimple way to prevent neural networks from overfitting[J].The Journal Machine Learning Research,2014,15(1):1929-1958.
[24]He Kai-ming,Zhang Xiang-yu,Ren Shao-qing,et al.Deep residual learning for image recognition[C]∥Proc of the2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:770-778.
[25]Szegedy C,Vanhoucke V,Ioffe S,et al.Rethinking the inception architecture for computer vision[C]∥Proc of the2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:2818-2826.
[26]Zheng Liang,Shen Li-yue,Tian Lu,et al.Scalable person reidentification:A benchmark[C]∥Proc of the 2015IEEEInternational Conference on Computer Vision(ICCV),2015:1116-1124.
[27]Li Wei,Zhao Rui,Xiao Tong,et al.DeepReID:Deep filter pairing neural network for person re-identification[C]∥Proc of the 2014IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2014:152-159.
[28]Ristani E,Solera F,Zou R S,et al.Performance measures and a data set for multi-target,multi-camera tracking[C]∥Proc of the European Conference on Computer Vision(EC-CV),2016:17-35.
[29]Ustinova E,Ganin Y,Lempitsky V S.Multi-region bilinear convolutional neural networks for person re-identification[C]∥Proc of the 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance(AVSS),2017:1-6.
[30]Zhang Li,Xiang Tao,Gong Shao-gang.Learning a discriminative null space for person re-identification[C]∥Proc of the 2016IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016:1239-1248.
[31]Barbosa I B,Cristani M,Caputo B,et al.Looking beyond appearances:Synthetic training data for deep CNNs in re-identification[J].Computer Vision and Image Understanding,2018,167:50-62.
[32]Lin Yu-tian,Zheng Liang,Zheng Zhe-dong,et al.Improving person re-identification by attribute and identity learning[J].arXiv preprint arXiv:1703.07220v2,2017.
[7]彭志勇,常发亮,刘洪彬,等.基于HSV模型和特征点匹配的行人重识别算法[J].光电子·激光,2015,26(8):1575-1582.
[8]黄新宇,许娇龙,郭纲,等.基于增强聚合通道特征的实时行人重识别[J].激光与光电子学进展,2017,54(9):113-121.
[9]孙金玉,王洪元,张继,等.基于块稀疏表示的行人重识别方法[J].计算机应用,2018,38(2):448-453.