Abstract：Pedestrian attributes usually refer to some of the external characteristics of pedestrians that can be observed, such as gender, age, clothing type, carrying objects, etc. As soft biological features of pedestrians, pedestrian attributes are very important for pedestrian detection and re-identification, and show great potential in intelligent video surveillance scenarios and video based business intelligence applications. Among the current multi-label classification methods of pedestrian attributes, two of them are mainly employed, one is based on handcrafted features and the other is based on the deep learning methods. However, the methods of handcrafted features are difficult to deal with complex real video surveillance scenes, results obtained in practical applications are not ideal. In this paper we used a deep convolutional network model with three convolutional layers and two full-connected layers. Using the Sigmoid cross-entropy loss function, the training platform was the Caffe deep learning framework, the dataset used was PETA containing 19,000 pedestrian images. Ten kinds of pedestrian attributes were trained and tested, and an average recognition accuracy of 85.2% was reached. After adding the positive sample proportional exponential factor to improve the loss function, the average recognition accuracy reached 89.2%, which significantly improved the performance of the network.
李亚鹏, 万遂人. 基于深度学习的行人属性多标签识别[J]. 中国生物医学工程学报, 2018, 37(4): 423-428.
Li Yapeng, Wan Suiren. Multi-Label Recognition of Pedestrian Attributes Based on Deep Learning. Chinese Journal of Biomedical Engineering, 2018, 37(4): 423-428.
[1] Jaha ES, Nixon MS. Analysing soft clothing biometrics for retrieval[C]// Biometric Authentication. Cham: Springer International Publishing, 2014: 234-245. [2] Dantcheva A, Singh A, Elia P, et al. Search pruning in video surveillance systems: Efficiency-reliability tradeoff[C]// 2011 IEEE International Conference on Computer Vision Workshops (ICCV Workshops). Barcelona: IEEE, 2011: 1356-1363. [3] Reid DA, Nixon MS, Stevenage SV. Soft biometrics; human identification using comparative descriptions [J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 36 (6): 1216-1228. [4] An Le, Chen Xiaojing, Kafai M, et al. Improving person re-identification by soft biometrics based reranking[C]// 2013 Seventh International Conference on Distributed Smart Cameras (ICDSC). Palm Springs: IEEE, 2013: 1-6. [5] Dantcheva A, Dugelay JL, Elia P. Person recognition using a bag of facial soft biometrics (BoFSB) [C]//2010 IEEE International Workshop on Multimedia Signal Processing (MMSP). Saint Malo: IEEE, 2010: 511-516. [6] Martinson E, Lawson E, Trafton G. Identifying people with soft-biometrics at fleet week[C]//Proceedings of the 8th ACM/IEEE international conference on Human-robot interaction. Tokyo: IEEE, 2013: 49-56. [7] Kumar N, Berg AC, Belhumeur PN, et al. Attribute and simile classifiers for face verification[C]//2009 IEEE 12th International Conference on Computer Vision. Kyoto: IEEE, 2009: 365-372. [8] Farenzena M, Bazzani L, Perina A, et al. Person re-identification by symmetry-driven accumulation of local features[C]//2010 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). San Francisco: IEEE, 2010: 2360-2367. [9] Ferrari V, Zisserman A. Learning visual attributes[C]//Advances in Neural Information Processing Systems. Vancouver: ACM, 2007: 433-440. [10] Siddiquie B, Feris RS, Davis LS. Image ranking and retrievalbased on multi-attribute queries[C]//2011 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Colorado Springs: IEEE, 2011: 801-808. [11] Kumar N, Berg AC, Belhumeur PN, et al. Describable visual attributes for face verification and image search[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(10): 1962-1977. [12] Layne R, Hospedales TM, Gong S, et al. Person reidentification by attributes[J]. BMVC, 2012, 2(3): 8-17. [13] Layne R, Hospedales TM, Gong S. Attributes-based re-identification[C]// Person Re-Identification. London: Springer, 2014: 93-117. [14] Zhu Jianqing, Liao Shengcai, Lei Zhen, et al. Pedestrian attribute classification in surveillance: Database and evaluation[C]//2013 IEEE International Conference on Computer Vision Workshops (ICCVW). Sydney: IEEE, 2013: 331-338. [15] Deng Yubin, Luo Ping, Loy CC, et al. Pedestrian attribute recognition at far distance[C]//Proceedings of the 22nd ACM International Conference on Multimedia. Orlando: ACM, 2014: 789-792. [16] Li Dangwei, Chen Xiaotang, Huang Kaiqi. Multi-attribute Learning for Pedestrian Attribute Recognition in Surveillance Scenarios[C]//2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR). Kuala Lumpur: IEEE, 2015: 111-115. [17] Zhu Jianqing, Liao Shengcai, Lei Zhen,et al. Multi-label convolutional neural network based pedestrian attribute classification [J]. Image and Vision Computing, 2017, 58: 224-229. [18] Fukui H, Yamashita T, Yamauchi Y, et al. Robust pedestrian attribute recognition for an unbalanced dataset using mini-batch training with rarity rate[C]// 2016 IEEE Intelligent Vehicles Symposium (IV). Gothenburg: IEEE, 2016: 322-327. [19] Levi G, Hassner T. Age and gender classification using convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. Boston: IEEE, 2015: 34-42. [20] Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25(2): 1097-1105. [21] Jia Yangqing, Shelhamer E, Donahue J, et al. Caffe: Convolutional architecture for fast feature embedding[C]//Proceedings of the 22nd ACM International Conference on Multimedia. Orlando: ACM, 2014: 675-678.
