Traffic Vehicle Detection Using Faster R-CNN: A Comparative Analysis of Backbone Architectures

Luqman  Hakim; Aria Hendrawan; Rofiatul Khoiriyah

doi:10.63158/IJAIS.v1.i1.5

Authors

Luqman Hakim Semarang University Author
Aria Hendrawan Semarang University Author
Rofiatul Khoiriyah Semarang University Author

DOI:

https://doi.org/10.63158/IJAIS.v1.i1.5

Keywords:

deep learning, Faster R-CNN, MobileNetV3, ResNet50, vehicle detection, computer vision

Abstract

Object detection is a crucial task in computer vision, where advanced deep learning models have shown significant improvements over traditional methods. In this study, the Faster R-CNN algorithm is applied to a traffic dataset containing six vehicle categories: Bus, Car, Motorcycle, Pick Up Car, Truck, and Truck Box. The novelty of the research lies in the comparison of four backbone architectures ResNet50, ResNet50V2, MobileNetV3 Large, and MobileNetV3 Large 320 evaluated for their performance in vehicle detection at IoU thresholds of 0.5 and 0.75. The results reveal that ResNet50 provided the best overall performance, achieving mAP scores of 0.966 at IoU 0.5 and 0.887 at IoU 0.75, offering a balanced trade-off between precision and recall. ResNet50V2 and MobileNetV3 Large also performed well, with mAP scores of 0.945 and 0.870 for ResNet50V2, and 0.969 and 0.843 for MobileNetV3 Large, respectively. However, MobileNetV3 Large 320 showed the lowest detection performance, with mAP scores of 0.857 at IoU 0.5 and 0.551 at IoU 0.75. These findings provide useful insights into the suitability of different architectures for vehicle detection tasks, particularly in traffic surveillance applications.

References

R. Padilla, S. L. Netto, and E. A. B. Da Silva, “A Survey on Performance Metrics for Object-Detection Algorithms,” in Int. Conf. Syst. Signals Image Process., vol. 2020-July, pp. 237–242, Jul. 2020, doi: 10.1109/IWSSIP48289.2020.9145130.

X. Wu, D. Sahoo, and S. C. H. Hoi, “Recent Advances in Deep Learning for Object Detection,” ACM Comput. Surv. (incomplete information, add more details).

H. Zhang and X. Hong, “Recent progresses on object detection: a brief review,” Multimed. Tools Appl., vol. 78, pp. 27809–27847, 2019, doi: 10.1007/s11042-019-07898-2.

M. Aria and C. Cuccurullo, “bibliometrix: An R-tool for comprehensive science mapping analysis,” J. Informetr., vol. 11, no. 4, pp. 959–975, Nov. 2017, doi: 10.1016/J.JOI.2017.08.007.

C. Janiesch, P. Zschech, and K. Heinrich, “Machine learning and deep learning,” Electron. Mark., vol. 31, no. 2, pp. 685–695, Apr. 2021, doi: 10.1007/s12525-021-00475-2.

S. Albahli, M. Nawaz, A. Javed, and A. Irtaza, “An improved Faster R-CNN model for handwritten character recognition,” Arab. J. Sci. Eng., vol. 46, no. 3, pp. 8509–8523, 2021, doi: 10.1007/s13369-021-05471-4.

S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 6, pp. 1137–1149, Jun. 2017, doi: 10.1109/TPAMI.2016.2577031.

J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., vol. 2016-December, pp. 779–788, Jun. 2016, doi: 10.1109/CVPR.2016.91.

W. Liu et al., “SSD: Single shot multibox detector,” in Lect. Notes Comput. Sci., vol. 9905, pp. 21–37, 2016, doi: 10.1007/978-3-319-46448-0_2.

W. Yang, Z. Li, C. Wang, and J. Li, “A multi-task Faster R-CNN method for 3D vehicle detection based on a single image,” Appl. Soft Comput., vol. 95, p. 106533, Oct. 2020, doi: 10.1016/J.ASOC.2020.106533.

S. Sinha and N. Gupta, “Computer-aided diagnosis of malaria through transfer learning using the ResNet50 backbone,” J. Med. Syst., vol. 36, no. 4, 2023.

A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang, Y. Zhu, R. Pang, V. Vasudevan, and Q. V. Le, “Searching for MobileNetV3,” in Proc. IEEE/CVF Int. Conf. Comput. Vis., pp. 1314–1324, 2019.

A. Hendrawan, R. Gernowo, O. D. Nurhayati, B. Warsito, and A. Wibowo, “Improvement Object Detection Algorithm Based on YoloV5 with BottleneckCSP,” in Proc. IEEE Int. Conf. Commun. Netw. Satellite (COMNETSAT), pp. 79–83, 2022, doi: 10.1109/COMNETSAT56033.2022.9994461.

S. Yang, D. Jiao, T. Wang, and Y. He, “Tire Speckle Interference Bubble Defect Detection Based on Improved Faster RCNN-FPN,” Sensors, vol. 22, no. 10, p. 3907, May 2022, doi: 10.3390/S22103907.