Abstract
This project makes use of a Convolutional Neural Network (CNN) to perform multi-class attribute recognition, in which this information is used to perform person re-identification (re-ID). From our research, we discovered that a trained CNN model performs better when given less attributes per image to focus on, as it decreases chances of error when making predictions of attributes of a person based on an image. Moreover, we found out that re-ID is done more effectively when a CNN is tasked to identify attributes that causes a person to stand out from others. Thus, salient attributes that can be clearly identified from cameras of different viewpoints are the most important attributes to focus on to perform re-ID effectively, while more common attributes can perform a filtering role in the re-ID problem. By modifying the Inception v3 model [1] for multi-label classification, the model is able to output probabilities for each attribute for every input image. Experiments on the PETA (PEdesTrian Attribute) dataset [2] has shown that the model performs better while recognising salient attributes only compared to recognising both common and salient attributes.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2818–2826).
Deng, Y., Luo, P., Loy, C. C., & Tang, X. (2014, November). Pedestrian attribute recognition at far distance. In Proceedings of the 22nd ACM international conference on Multimedia (pp. 789–792). ACM.
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., et. al. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1–9).
Xu, M., Tang, Z., Yao, Y., Yao, L., Liu, H., & Xu, J. (2017). Deep learning for person reidentification using support vector machines. Advances in Multimedia, 2017.
Li, D., Chen, X., & Huang, K. (2015, November). Multi-attribute learning for pedestrian attribute recognition in surveillance scenarios. In 2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR) (pp. 111–115). IEEE.
Li, D., Zhang, Z., Chen, X., Ling, H., & Huang, K. (2016). A richly annotated dataset for pedestrian attribute recognition. arXiv:1603.07054 (arXiv preprint).
Acknowledgements
The authors, Poh Say Keong and Wen Jun Gao Calvin, would like to thank mentor Dr. Shen Bing Quan from DSO National Laboratories and teacher mentor Mr. Lee Wei Keong from Dunman High School for their support and guidance.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix: Code to Calculate Precision and Recall of Each Attribute
Appendix: Code to Calculate Precision and Recall of Each Attribute
Rights and permissions
Copyright information
© 2019 Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Gao, W.J.C., Keong, P.S., Shen, B. (2019). Human Attribute Classification for Re-identification Across Non-overlapping Cameras. In: Guo, H., Ren, H., Bandla, A. (eds) IRC-SET 2018. Springer, Singapore. https://doi.org/10.1007/978-981-32-9828-6_7
Download citation
DOI: https://doi.org/10.1007/978-981-32-9828-6_7
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-32-9827-9
Online ISBN: 978-981-32-9828-6
eBook Packages: Biomedical and Life SciencesBiomedical and Life Sciences (R0)