OBJECT DETECTION SYSTEMS FOR DETERMINING CAR MODELS AND BODY COLOR: COMPREHENSIVE COMPARATIVE ANALYSIS AND PRACTICAL INTEGRATION BASED ON REAL IMAGES UNDER UZBEKISTAN CONDITIONS

Davronov Rifkat Rakhimovich; Misirov Farxod Abdulla ugli

Authors

Davronov Rifkat Rakhimovich V.I.Romanovskiy Institute of Mathematics, Uzbekistan Academy of Sciences
Misirov Farxod Abdulla ugli V.I.Romanovskiy Institute of Mathematics, Uzbekistan Academy of Sciences

Keywords:

CCTV, object detection, vehicle model, body color estimation, YOLOv8-l, DETR (ResNet-101), CenterNet, CVAT, COCO/YOLO, LAB, HSV, F1–Confidence, real time, multi-camera, ONNX, TensorRT

Abstract

This paper presents a practical solution for real-time recognition of vehicle model and body color on real-world CCTV streams in Uzbekistan. The dataset was annotated in CVAT and cleaned; to ensure fair evaluation, we applied a camera-wise train/val/test split (preventing cross-camera leakage). On the same dataset, we fine-tuned YOLOv8-l, DETR (ResNet-101), and CenterNet, and analyzed the quality–speed trade-off from a deployment perspective. Using detector-produced bounding boxes as ROIs, we integrated a fast color module into the pipeline based on a combined LAB+HSV approach. For inference threshold selection, we employed a practical method grounded in the F1–Confidence curve. The system operates robustly with multi-camera RTSP streams, supports auto-reconnect and resource monitoring (CPU/RAM/IO and, when available, GPU), and is ready for ONNX/TensorRT export. The results demonstrate practical applicability in traffic monitoring, security, and intelligent video surveillance systems.

References

1. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. In: Proceedings of CVPR, 779–788.

2. Bochkovskiy, A., Wang, C.-Y., & Liao, H.-Y. M. (2020). YOLOv4: Optimal Speed and Accuracy of Object Detection. arXiv:2004.10934.

3. Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., & Zagoruyko, S. (2020). End-to-End Object Detection with Transformers (DETR). In: Proceedings of ECCV, 213–229.

4. Duan, K., Bai, S., Xie, L., Qi, H., Huang, Q., & Tian, Q. (2019). CenterNet: Keypoint Triplets for Object Detection. In: Proceedings of ICCV, 6569–6578.

5. Tan, M., & Le, Q. V. (2019). EfficientDet: Scalable and Efficient Object Detection. arXiv:1911.09070.

6. Lin, T.-Y., Maire, M., Belongie, S., et al. (2014). Microsoft COCO: Common Objects in Context. In: Proceedings of ECCV, 740–755.

7. Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollár, P. (2017). Focal Loss for Dense Object Detection (RetinaNet). In: Proceedings of ICCV, 2980–2988.

8. Rezatofighi, H., Tsoi, N., Gwak, J., et al. (2019). Generalized Intersection over Union: A Metric and a Loss for Bounding Boxes. In: Proceedings of CVPR, 658–666.

9. Talib, M., Al-Noori, A. H. Y., & Suad, J. (2024). YOLOv8-CAB: Improved YOLOv8 for Real-time Object Detection. Karbala International Journal of Modern Science