Abstract
In this paper, we improve the accuracy of person re-identification in images obtained from distributed video surveillance systems by choosing activation functions for convolutional neural networks. The most popular activation functions used for object detection, namely, ReLU, Leaky-ReLU, PReLU, RReLU, ELU, SELU, GELU, Swish, and Mish, are analyzed based on the following metrics: Rank1, Rank5, Rank10, mAP, and training time. For feature extraction, ResNet-50, DenseNet-121, and DarkNet-53 architectures are employed. The experimental study is carried out on open datasets Market1501 and PolReID. The accuracy of person re-identification is assessed after thrice-repeated training and testing with different activation functions, neural network architectures, and datasets by averaging the values of the selected metrics.
Similar content being viewed by others
REFERENCES
Ye, S., Bohush, R.P., Chen, H., et al., Person tracking and reidentification for multicamera indoor video surveillance systems, Pattern Recognit. Image Anal., 2020, vol. 30, pp. 827–837. https://doi.org/10.1134/S1054661820040136
Porrello, A., Bergamini, L., and Calderara, S., Robust re-identification by multiple views knowledge distillation, 2020.
Huang, G., Liu, Z., and Weinberger, K.Q., Densely connected convolutional networks, Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2017, pp. 2261–2269.
Wang, G., Lai, J., Huang, P., and Xie, X., Spatial-temporal person re-identification, 2019.
Mao, S., Zhang, S., and Yang, M., Resolution-invariant person re-identification, 2019.
Redmon, J. and Farhadi, A., YOLOv3: An incremental improvement, 2018.
Nair, V. and Hinton, G.E., Rectified linear units improve restricted Boltzmann machines, Proc. ICML, 2010, pp. 807–814.
Maas, A.L., Rectifier non linearities improve neural network acoustic models, Proc. ICML, 2013, vol. 30.
Xu, B., Wang, N., Chen, T., and Li, M., Empirical evaluation of rectified activations in convolutional network, 2015.
Clevert, D., Unterthiner, T., and Hochreiter, S., Fast and accurate deep network learning by exponential linear units (ELUs), 2016.
Klambauer, G., Unterthiner, T., Mayr, A., and Hochreiter, S., Self-normalizing neural networks, 2017.
Hendrycks, D. and Gimpel, K., Bridging nonlinearities and stochastic regularizers with Gaussian error linear units, 2016.
Ramachandran, P., Zoph, B., and Le, Q.V., Swish: A self-gated activation function, 2017.
Misra, D., Mish: A self regularized non-monotonic neural activation function, 2019.
Zheng, L., Shen, L., Tian, L., Wang, Sh., Wang, J., and Tian, Q., Scalable person re-identification: A benchmark, Proc. IEEE Int. Conf. Computer Vision (ICCV), 2015, pp. 1116–1124.
Ihnatsyeva, S., Bohush, R., and Ablameyko, S., Joint dataset for CNN-based person re-identification, Proc. 15th Int. Conf. Pattern Recognition and Information Processing (PRIP), Minsk, Minsk: Ob"edinennyi Inst. Probl. Inf. Nats. Akad. Nauk Belarusi, 2021, pp. 33–37.
Ihnatsyeva, S. and Bohush, R., PolReID. https://github.com/SvetlanaIgn/PolReID.
Bochkovskiy, A., Wang, Ch.-Y., and Liao, H.-Y.M., YOLOv4: Optimal speed and accuracy of object detection, 2020.
GitHub, Person reID baseline pytorch. https://github.com/layumi/Person_reID_baseline_pytorch.
ACKNOWLEDGMENTS
This work is partially supported by the National High-end Foreign Experts Program (G2021016028L and G2021016002L) and Zhejiang Shuren University Basic Scientific Research Special Funds.
Author information
Authors and Affiliations
Corresponding authors
Ethics declarations
The authors declare that they have no conflicts of interest.
Additional information
Translated by Yu. Kornienko
Rights and permissions
About this article
Cite this article
Chen, H., Ihnatsyeva, S., Bohush, R. et al. Choice of Activation Function in Convolutional Neural Networks for Person Re-Identification in Video Surveillance Systems. Program Comput Soft 48, 312–321 (2022). https://doi.org/10.1134/S0361768822050036
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1134/S0361768822050036