Otimização de pipeline multicritério para detecção de pessoas em dispositivos ultracompactos de borda
Resumo
A computação de borda consolidou-se como estratégia fundamental para aplicações de visão computacional que exigem baixa latência e controle de privacidade. Contudo, dispositivos de borda enfrentam restrições de consumo energético e custo operacional, dificultando a escolha da configuração ideal. Este artigo propõe um pipeline multicritério para dispositivos ultracompactos e compara reCamera, Jetson Xavier NX e desktop, usando os datasets MOT17 e Pedestrian com wattímetro externo e métricas por faixas DORI. Embora o desktop apresente os melhores resultados, a reCamera opera com 1,9 W constante, até 59× menor que o desktop e 50× mais barata, sendo a melhor alternativa para ambientes com infraestrutura elétrica restrita. O pipeline indica 720p a 5 FPS como configuração ótima para cenas de baixa densidade e 480p a 30 FPS para implantações em tempo real.Referências
Alqahtani, D. K., Cheema, A., and Toosi, A. N. (2024). Benchmarking deep learning models for object detection on edge computing devices. arXiv preprint arXiv:2409.16808.
Ammar, O., Benyahya, S., and El Fkihi, S. (2024). Energy-efficient edge AI inference for real-time vision applications. Sensors.
Axis Communications (2023). Densidade de pixels e DORI: atender aos requisitos operacionais no vídeo em rede. Technical Report T10176489/PT/M2.3/2305, Axis Communications AB.
Bonomi, F., Milito, R., Zhu, J., and Addepalli, S. (2012). Fog computing and its role in the Internet of Things. In Proceedings of the First Edition of the MCC Workshop on Mobile Cloud Computing, MCC ’12, pages 13–16. ACM.
Cao, K., Liu, Y., Meng, G., and Sun, Q. (2023). A survey of AI-enabled edge computing: architectures, applications and challenges. IEEE Access, 11:10240–10257.
Choe, C., Choe, M., and Jung, S. (2023). Run your 3D object detector on NVIDIA Jetson platforms: a benchmark analysis. Sensors, 23(8):4005.
Dendorfer, P., Osep, A., Milan, A., Schindler, K., Cremers, D., Reid, I., Roth, S., and Leal-Taixé, L. (2021). Motchallenge: A benchmark for single-camera multiple target tracking: P. dendorfer et al. International Journal of Computer Vision, 129(4):845–881.
Dendorfer, P., Rezatofighi, H., Milan, A., Shi, J., Cremers, D., Reid, I., Roth, S., Schindler, K., and Leal-Taixé, L. (2020). Mot20: A benchmark for multi object tracking in crowded scenes. arXiv preprint arXiv:2003.09003.
Desislavov, R., Martínez-Plumed, F., and Hernández-Orallo, J. (2023). Compute and energy consumption trends in deep learning inference. Patterns, 4(1):100i.
Dhar, S. (2012). From outsourcing to cloud computing: evolution of IT services. Management Research Review, 35(8):664–675.
Ferraz, L. et al. (2025). A performance analysis of YOLO-based models for quantized object detection on edge devices. arXiv preprint arXiv:2502.15737.
González, M. L. et al. (2025). Deep learning inference on edge: a preliminary device comparison. In Intelligent Data Engineering and Automated Learning – IDEAL 2024, volume 15346 of Lecture Notes in Computer Science, pages 247–258. Springer.
Hu, Y., Liu, H., Pfeiffer, M., and Delbruck, T. (2016). Dvs benchmark datasets for object tracking, action recognition, and object recognition. Frontiers in neuroscience, 10:405.
International Electrotechnical Commission (2015). IEC EN62676-4: video surveillance systems for use in security applications – Part 4: application guidelines. International standard, IEC.
Karthika, N. J. and Saravanan, C. (2020). Addressing false positives in pedestrian detection. In International Conference on Electronic Systems and Intelligent Computing (ESIC 2020).
Lema, D. G., Usamentiaga, R., and García, D. F. (2024). Quantitative comparison and performance evaluation of deep learning-based object detection models on edge computing devices. Integration, 95:102127.
Liang, S., Wu, H., Zhen, L., Hua, Q., Garg, S., Kaddoum, G., Hassan, M. M., and Yu, K. (2022). Edge YOLO: real-time intelligent object detection system based on edge-cloud cooperation in autonomous vehicles. IEEE Transactions on Intelligent Transportation Systems, 23(12):25345–25360.
Liu, Y., Zhang, H., et al. (2023). Q-YOLO: efficient inference for real-time object detection. arXiv preprint arXiv:2307.04816.
Mao, Y., You, C., Zhang, J., Huang, K., and Letaief, K. B. (2024). Green edge AI: a contemporary survey. Proceedings of the IEEE, 112(7):880–918.
Milan, A., Leal-Taixé, L., Reid, I., Roth, S., and Schindler, K. (2016). MOT16: a benchmark for multi-object tracking. arXiv preprint arXiv:1603.00831.
Mittal, P. (2024). A comprehensive survey of deep learning-based lightweight object detection models for edge devices. Artificial Intelligence Review, 57(9):242.
Seeed Studio (2025). reCamera – intelligent edge vision device. Página do produto. Acesso em: 19 mar. 2026.
Shi, W., Cao, J., Zhang, Q., Li, Y., and Xu, L. (2016). Edge computing: vision and challenges. IEEE Internet of Things Journal, 3(5):637–646.
Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., and Chen, J. (2023). DETRs beat YOLOs on real-time object detection. arXiv preprint arXiv:2304.08069.
Ammar, O., Benyahya, S., and El Fkihi, S. (2024). Energy-efficient edge AI inference for real-time vision applications. Sensors.
Axis Communications (2023). Densidade de pixels e DORI: atender aos requisitos operacionais no vídeo em rede. Technical Report T10176489/PT/M2.3/2305, Axis Communications AB.
Bonomi, F., Milito, R., Zhu, J., and Addepalli, S. (2012). Fog computing and its role in the Internet of Things. In Proceedings of the First Edition of the MCC Workshop on Mobile Cloud Computing, MCC ’12, pages 13–16. ACM.
Cao, K., Liu, Y., Meng, G., and Sun, Q. (2023). A survey of AI-enabled edge computing: architectures, applications and challenges. IEEE Access, 11:10240–10257.
Choe, C., Choe, M., and Jung, S. (2023). Run your 3D object detector on NVIDIA Jetson platforms: a benchmark analysis. Sensors, 23(8):4005.
Dendorfer, P., Osep, A., Milan, A., Schindler, K., Cremers, D., Reid, I., Roth, S., and Leal-Taixé, L. (2021). Motchallenge: A benchmark for single-camera multiple target tracking: P. dendorfer et al. International Journal of Computer Vision, 129(4):845–881.
Dendorfer, P., Rezatofighi, H., Milan, A., Shi, J., Cremers, D., Reid, I., Roth, S., Schindler, K., and Leal-Taixé, L. (2020). Mot20: A benchmark for multi object tracking in crowded scenes. arXiv preprint arXiv:2003.09003.
Desislavov, R., Martínez-Plumed, F., and Hernández-Orallo, J. (2023). Compute and energy consumption trends in deep learning inference. Patterns, 4(1):100i.
Dhar, S. (2012). From outsourcing to cloud computing: evolution of IT services. Management Research Review, 35(8):664–675.
Ferraz, L. et al. (2025). A performance analysis of YOLO-based models for quantized object detection on edge devices. arXiv preprint arXiv:2502.15737.
González, M. L. et al. (2025). Deep learning inference on edge: a preliminary device comparison. In Intelligent Data Engineering and Automated Learning – IDEAL 2024, volume 15346 of Lecture Notes in Computer Science, pages 247–258. Springer.
Hu, Y., Liu, H., Pfeiffer, M., and Delbruck, T. (2016). Dvs benchmark datasets for object tracking, action recognition, and object recognition. Frontiers in neuroscience, 10:405.
International Electrotechnical Commission (2015). IEC EN62676-4: video surveillance systems for use in security applications – Part 4: application guidelines. International standard, IEC.
Karthika, N. J. and Saravanan, C. (2020). Addressing false positives in pedestrian detection. In International Conference on Electronic Systems and Intelligent Computing (ESIC 2020).
Lema, D. G., Usamentiaga, R., and García, D. F. (2024). Quantitative comparison and performance evaluation of deep learning-based object detection models on edge computing devices. Integration, 95:102127.
Liang, S., Wu, H., Zhen, L., Hua, Q., Garg, S., Kaddoum, G., Hassan, M. M., and Yu, K. (2022). Edge YOLO: real-time intelligent object detection system based on edge-cloud cooperation in autonomous vehicles. IEEE Transactions on Intelligent Transportation Systems, 23(12):25345–25360.
Liu, Y., Zhang, H., et al. (2023). Q-YOLO: efficient inference for real-time object detection. arXiv preprint arXiv:2307.04816.
Mao, Y., You, C., Zhang, J., Huang, K., and Letaief, K. B. (2024). Green edge AI: a contemporary survey. Proceedings of the IEEE, 112(7):880–918.
Milan, A., Leal-Taixé, L., Reid, I., Roth, S., and Schindler, K. (2016). MOT16: a benchmark for multi-object tracking. arXiv preprint arXiv:1603.00831.
Mittal, P. (2024). A comprehensive survey of deep learning-based lightweight object detection models for edge devices. Artificial Intelligence Review, 57(9):242.
Seeed Studio (2025). reCamera – intelligent edge vision device. Página do produto. Acesso em: 19 mar. 2026.
Shi, W., Cao, J., Zhang, Q., Li, Y., and Xu, L. (2016). Edge computing: vision and challenges. IEEE Internet of Things Journal, 3(5):637–646.
Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., and Chen, J. (2023). DETRs beat YOLOs on real-time object detection. arXiv preprint arXiv:2304.08069.
Publicado
25/05/2026
Como Citar
SANTOS, João B. A.; REGO, Paulo A. L..
Otimização de pipeline multicritério para detecção de pessoas em dispositivos ultracompactos de borda. In: WORKSHOP DE COMPUTAÇÃO URBANA (COURB), 10. , 2026, Praia do Forte/BA.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 197-210.
ISSN 2595-2706.
DOI: https://doi.org/10.5753/courb.2026.24093.
