Vehicle Detection and Tracking using EfficientDet and BoT-SORT

Abstract

With the rapid growth of urban traffic, there is an increasing need for efficient and automated monitoring systems for traffic management. In this work, a vehicle detection and tracking system is proposed using the EfficientDet object detection model combined with the BoT-SORT tracking algorithm. The system processes traffic surveillance video frames to detect vehicles and assign unique tracking IDs, ensuring continuity across consecutive frames. EfficientDet provides a good balance between detection accuracy and computational efficiency, while BoT-SORT improves tracking performance by maintaining consistent object identities even under dynamic traffic conditions. Experimental results show that the proposed approach is able to detect and track multiple vehicles effectively in real-world scenarios. The system demonstrates reliable performance and can be used for practical traffic monitoring applications.

Introduction

Traditional traffic surveillance methods rely on manual observation, which is inefficient, error-prone, and unsuitable for large-scale deployment. With advances in deep learning and computer vision, automated systems have become more effective for detecting and tracking vehicles in real time.

The proposed system combines EfficientDet for vehicle detection and BoT-SORT for multi-object tracking. EfficientDet extracts features using an EfficientNet backbone and a Bi-directional Feature Pyramid Network (BiFPN), enabling accurate and efficient detection of vehicles such as cars, buses, and trucks. BoT-SORT then assigns unique IDs to each detected vehicle and maintains consistent tracking across frames using both motion and appearance information, even in complex traffic conditions.

The system processes video inputs frame by frame using OpenCV, detects vehicles in each frame, tracks them across time, and generates movement trajectories. These trajectories represent vehicle paths and can support further traffic analysis, such as flow monitoring and behavior prediction.

The methodology includes frame extraction, object detection, tracking, ID assignment, and trajectory generation, producing annotated video outputs that visualize real-time traffic movement.

The literature review highlights the evolution of object detection models from Faster R-CNN to YOLO and EfficientDet, as well as tracking methods like SORT, DeepSORT, ByteTrack, and BoT-SORT. Among these, EfficientDet offers a balance between accuracy and efficiency, while BoT-SORT improves tracking stability in dynamic environments.

Conclusion

In this paper, a vehicle detection and tracking system based on an EfficientDet-inspired framework has been presented for real-time traffic monitoring applications. The proposed approach integrates deep learning-based object detection with a tracking mechanism to accurately identify and monitor multiple vehicles across video frames. The experimental results demonstrate that the system is capable of detecting vehicles such as cars, buses, trucks, and motorcycles with good accuracy under moderate traffic conditions. The system maintains consistent performance across frames, as observed from the vehicle count analysis, and achieves near real-time processing speed, as indicated by the FPS evaluation. The graphical analysis highlights the relationship between detection load and processing speed, where an increase in the number of detected vehicles leads to a slight reduction in FPS. However, the system maintains stable performance within acceptable limits, demonstrating its robustness and suitability for practical deployment. Overall, the proposed system provides an effective and efficient solution for vehicle detection and tracking in traffic surveillance scenarios. The combination of detection accuracy and real-time capability makes it a promising approach for intelligent transportation systems and smart city applications.

References

[1] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” in Proc. Advances in Neural Information Processing Systems (NeurIPS), 2015. [2] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, and S. Reed, “SSD: Single Shot MultiBox Detector,” in Proc. European Conf. Computer Vision (ECCV), 2016. [3] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2016. [4] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” 2020. [5] C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, “YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art,” 2022. [6] M. Tan, R. Pang, and Q. V. Le, “EfficientDet: Scalable and Efficient Object Detection,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2020. [7] M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” in Proc. International Conf. Machine Learning (ICML), 2019. [8] Y. Zhang, X. Wang, and H. Chen, “Vehicle detection using EfficientDet in traffic surveillance,” IEEE Access, 2023. [9] R. Kumar, S. Singh, and A. Sharma, “EfficientDet-based vehicle detection for real-time applications,” Journal of Intelligent Transportation Systems, 2024. [10] A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple Online and Realtime Tracking (SORT),” in Proc. IEEE Int. Conf. Image Processing, 2016. [11] N. Wojke, A. Bewley, and D. Paulus, “Simple Online and Realtime Tracking with a Deep Association Metric (DeepSORT),” 2017. [12] Y. Zhang, P. Sun, Y. Jiang, and D. Yuan, “ByteTrack: Multi-Object Tracking by Associating Every Detection Box,” 2022. [13] N. Aharon, R. Orfaig, and B. Z. Bobrovsky, “BoT-SORT: Robust Associations Multi-Pedestrian Tracking,” 2022. [14] G. Amato, F. Carrara, F. Falchi, C. Gennaro, and C. Vairo, “Deep learning for decentralized parking lot occupancy detection,” Expert Systems with Applications, 2017. [15] X. Yin, X. Wang, and Z. Liu, “Deep learning-based parking detection system,” IEEE Access, 2019. [16] Z. Carion et al., “End-to-End Object Detection with Transformers (DETR),” in Proc. ECCV, 2020. [17] H. Zhang, Y. Li, and Z. Wang, “Multi-view vehicle detection and tracking for traffic monitoring,” IEEE Transactions on Intelligent Transportation Systems, 2023. [18] L. Breiman, “Random Forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001. [19] C. M. Bishop, Pattern Recognition and Machine Learning. Springer, 2006. [20] Ultralytics, “YOLOv8 Documentation,” 2023.

Copyright

Copyright © 2026 Mrs. Prachi Sodegaonkar, Dr. R. A. Khan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.