Automated Large-Scale Labeling of Brazilian Urban Traffic Images with Model Ensembles

• Carlos Loureiro UEA
• William Harada UEA
• Elloá B. Guedes UEA
• Carlos Maurício S. Figueiredo UEA

DOI: https://doi.org/10.5753/eniac.2026.29869

Resumo

Custom dataset creation remains a time-consuming bottleneck for training detection models. To advance urban mobility analysis in Brazil, we propose an autonomous, large-scale labeling pipeline using a vision model ensemble (Grounding DINO, OWLv2, YOLOv11) to annotate traffic surveillance images. Training the RT-DETR architecture on the resulting dataset yielded an mAP@0.5:0.95 of 0.5265 (F₁: 73.59%) on synthetic labels, and 0.4651 (F₁: 74.02%) on human-supervised ground truth. This strong real-world generalization offers a highly scalable alternative to manual annotation for intelligent urban traffic monitoring.

Referências

Barbado, L., Santos, L., Matrakas, M., and Moreira, J. (2022). Aplicação da rede convolucional yolo para análise de fluxo de veículos. In Anais do XIX Congresso Latino-Americano de Software Livre e Tecnologias Abertas, pages 43–49, Porto Alegre, RS, Brasil. SBC.

Brasil (2025). Ministério dos transportes – secretaria nacional de trânsito – frota de veículos. Disponível em [link]. Acessado em: 15 maio 2026.

Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., and Zagoruyko, S. (2020). End-to-end object detection with transformers. In European Conference on Computer Vision, pages 213–229. Springer.

Carvalho, R., Guedes, E., and Figueiredo, C. (2022). Detecção inteligente de falhas em pavimentações asfálticas com redes neurais convolucionais regionais. In Anais do III Workshop Brasileiro de Cidades Inteligentes, pages 119–130, Porto Alegre, RS, Brasil. SBC.

Chan, P. H., Li, B., Baris, G., Sadiq, Q., and Donzella, V. (2024). The inconvenient truth of ground truth errors in automotive datasets and dnn-based detection. Data-Centric Engineering, 5:e34.

Freire, A., de S. Silva, L. H., de Andrade, J. V., Azevedo, G. O., and Fernandes, B. J. (2024). Beyond clean data: Exploring the effects of label noise on object detection performance. Knowledge-Based Systems, 304:112544.

Júnior, P. S. D. S. F. and Guedes, E. B. (2023). Detecção de semáforos em ambientes urbanos para cidades inteligentes. In Procedings do XVI Simpósio Brasileiro de Automação Inteligente / X Simpósio Brasileiro de Sistemas Elétricos, SBAI2023 / SBSE2023. SBA Sociedade Brasileira de Automática.

Liu, S., Zeng, Z., Ren, T., Li, F., Zhang, H., Yang, J., Jiang, Q., Li, C., Yang, J., Su, H., Zhu, J., and Zhang, L. (2024). Grounding dino: Marrying dino withnbsp;grounded pre-training fornbsp;open-set object detection. In Computer Vision – ECCV 2024: 18th European Conference, Milan, Italy, September 29–October 4, 2024, Proceedings, Part XLVII, page 38–55, Berlin, Heidelberg. Springer-Verlag.

Loureiro, C., Guedes, E., and Figueiredo, C. (2025). Monitoramento inteligente de tempo real para tráfego urbano em cidades brasileiras. In Anais do XVII Simpósio Brasileiro de Computação Ubíqua e Pervasiva, pages 41–50. SBC.

Minderer, M., Gritsenko, A., and Houlsby, N. (2023). Scaling open-vocabulary object detection. In Proceedings of the 37th International Conference on Neural Information Processing Systems, NIPS ’23, Red Hook, NY, USA. Curran Associates Inc.

Myagmar-Ochir, Y. and Kim, W. (2023). A survey of video surveillance systems in smart city. Electronics, 12(17):3567.

Redmon, J., Divvala, S., Girshick, R. B., and Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 779–788.

Santos, A. M., Bastos-Filho, C. J. A., Maciel, A. M. A., and Lima, E. (2020). Counting vehicle with high-precision in Brazilian roads using YOLOv3 and Deep SORT. In 2020 33rd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI). IEEE.

Santos, D., Claro, D. B., and Gondim, J. (2023). Monitoring vehicle plate detection in brazilian universities. In Proceedings of the XIX Brazilian Symposium on Information Systems, SBSI ’23, page 268–275. ACM.

Solovyev, R., Wang, W., and Gabruseva, T. (2021). Weighted boxes fusion: Ensembling boxes from different object detection models. Image and Vision Computing, 107:104117.

Souza, A. N., Vassallo, R., and Abling, A. (2025). Reidentificação espaço-temporal de veículos para sistemas inteligentes de tráfego. In Anais Estendidos do XLIII Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos, pages 370–375. SBC.

Velesaca, H. O., Mero, A., Singh, A., and Sappa, A. (2026). The hidden cost of poor annotations: How label quality affects camouflaged object detection performance. In Bebis, G., Ye, J., Wang, Y., Konaković Luković, M., Kalantari, N. K., Cho, I., Yang, Y., Dimara, E., and Brehmer, M., editors, Advances in Visual Computing, pages 285–296, Cham. Springer Nature Switzerland.

Vijayakumar, A. and Vairavasundaram, S. (2024). YOLO-based object detection models: A review and its applications. Multimed. Tools Appl.

Wu, T. (2024). Reducing the cost of test data labelling for deep-learning systems: An empirical study. Master’s thesis, School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, Ontario, Canada. Disponível em DOI: 10.20381/ruor-30788. Acesso em: 15 de maio de 2026.

Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., and Chen, J. (2024). DETRs beat YOLOs on real-time object detection. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16965–16974. IEEE.