반도체디스플레이기술학회지 (Journal of the Semiconductor & Display Technology)

한국반도체디스플레이기술학회 (The Korean Society Of Semiconductor & Display Technology)
권호 표지

백본 네트워크에 따른 사람 속성 검출 모델의 성능 변화 분석

Analyzing DNN Model Performance Depending on Backbone Network

  • 박천수 (성균관대학교 컴퓨터교육과)
  • Chun-Su Park (Computer Education, Sungkyunkwan University)
  • 투고 : 2023.06.15
  • 심사 : 2023.06.21
  • 발행 : 2023.06.30
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

Recently, with the development of deep learning technology, research on pedestrian attribute recognition technology using deep neural networks has been actively conducted. Existing pedestrian attribute recognition techniques can be obtained in such a way as global-based, regional-area-based, visual attention-based, sequential prediction-based, and newly designed loss function-based, depending on how pedestrian attributes are detected. It is known that the performance of these pedestrian attribute recognition technologies varies greatly depending on the type of backbone network that constitutes the deep neural networks model. Therefore, in this paper, several backbone networks are applied to the baseline pedestrian attribute recognition model and the performance changes of the model are analyzed. In this paper, the analysis is conducted using Resnet34, Resnet50, Resnet101, Swin-tiny, and Swinv2-tiny, which are representative backbone networks used in the fields of image classification, object detection, etc. Furthermore, this paper analyzes the change in time complexity when inferencing each backbone network using a CPU and a GPU.

키워드

참고문헌

  1. Y. Hu, et al. "A More Efficient Approach for Pedestrian Attribute Recognition," IEEE International Joint Conference on Biometrics (IJCB), pp. 1-8, 2022. 
  2. X. Wang, et al. "Pedestrian attribute recognition: A survey," Pattern Recognition, Vol. 121, pp. 108220, 2022. 
  3. Y. Deng, P. Luo, C.C. Loy, X. Tang, "Pedestrian attribute recognition at far distance," ACM Multimedia, pp. 789-792, 2014. 
  4. M. Tan and Q. Le, "Efficientnet: Rethinking model scaling for convolutional neural networks," Image Computing & Machine Learning (ICML), pp. 6105-6114, 2019. 
  5. Y. Chen, S. He, Z. Tan, C. Han, G. Han, J. Qin, "Age estimation via attribute-region association," Neurocomputing, Vol. 367, pp. 346-356, 2019.  https://doi.org/10.1016/j.neucom.2019.08.034
  6. J. Wang, X. Zhu, S. Gong, and W. Li, "Attribute recognition by joint recurrent learning of context and correlation," International Conference on Computer Vision (ICCV), 2017. 
  7. D. Li, X. Chen, and K. Huang, "Multi-attribute learning for pedestrian attribute recognition in surveillance scenarios," Asian Conference on Pattern Recognition (ACPR), pp. 111-115, 2015. 
  8. X. Liu, H. Zhao, M. Tian, L. Sheng, J. Shao, S. Yi, J. Yan, and X. Wang, "Hydraplus-net: Attentive deep features for pedestrian analysis," International Conference on Computer Vision (ICCV), 2017. 
  9. D. Weng, Z. Tan, L. Fang, and G. Guo, "Exploring attribute localization and correlation for pedestrian attribute recognition," Neurocomputing, Vol. 531, pp. 140-150, 2023.  https://doi.org/10.1016/j.neucom.2023.02.019
  10. J. Jia, et al. "Rethinking of pedestrian attribute recognition: A reliable evaluation under zero-shot pedestrian identity setting," arXiv preprint arXiv:2107.03576, 2021. 
  11. S. Targ, A. Diogo, and K. Lyman. "Resnet in resnet: Generalizing residual architectures," arXiv preprint arXiv:1603.08029, 2016. 
  12. Z. Wu, C. Shen, and A. D. Hengel, "Wider or deeper: Revisiting the resnet model for visual recognition," Pattern Recognition, Vol. 90, pp. 119-133, 2019.  https://doi.org/10.1016/j.patcog.2019.01.006
  13. T. He, et al. "Bag of tricks for image classification with convolutional neural networks," IEEE/CVF conference on computer vision and pattern recognition, pp. 558-567, 2019. 
  14. Z. Liu, et al. "Swin transformer: Hierarchical vision transformer using shifted windows," IEEE/CVF international conference on computer vision, pp. 10012-10022, 2021. 
  15. Z. Liu, et al. "Swin transformer v2: Scaling up capacity and resolution," IEEE/CVF conference on computer vision and pattern recognition, pp. 12009-1019, 2022. 
  16. Y. Deng, P. Luo, C. C. Loy, and X. Tang, "Pedestrian attribute recognition at far distance," ACM Int. Conf. Multimedia, pp. 789-792, 2014. 
  17. I. N. Junejo and N. Ahmed, "A multi-branch separable convolution neural network for pedestrian attribute recognition," Heliyon, Vol. 6, No. 3, e03563, 2020. 
  18. X. Liu, H. Zhao, M. Tian, L. Sheng, J. Shao, S. Yi, J. Yan, and X. Wang, "Hydraplus-net: Attentive deep features for pedestrian analysis," IEEE Int. Conf. Comput. Vis., pp. 350-359, 2017. 
  19. D. Li, Z. Zhang, X. Chen, and K. Huang, "A richly annotated pedestrian dataset for person retrieval in real surveillance scenarios," IEEE Trans. Image Process., Vol. 28, No. 4, pp. 1575-1590, 2018.  https://doi.org/10.1109/TIP.2018.2878349
  20. Y. Lin, L. Zheng, Z. Zheng, Y. Wu, Z. Hu, C. Yan, and Y. Yang, "Improving person re-identification by attribute and identity learning," Pattern Recognition, Vol. 95, pp. 151-161, 2019.  https://doi.org/10.1016/j.patcog.2019.06.006
  21. C. S. Park, "Performance Analysis of DNN inference using OpenCV Built in CPU and GPU Functions," Journal of the Semiconductor & Display Technology, Vol. 21, No. 1, pp. 75-78, 2022.