RESOLUTION-DEPENDENT LATENCY AND PARETO EFFICIENCY OF FACE DETECTORS FOR REAL-TIME EDGE BIOMETRIC SYSTEMS
DOI:
https://doi.org/10.47390/ydif-y2026v2i19/n06Keywords:
inference latency; input resolution; edge computing; real-time biometrics; Pareto efficiency; frame-rate budget; YuNet; Dlib.Abstract
The computational cost of face detection grows with the resolution of the video stream and often determines whether a biometric system can run in real time on edge hardware. This study analyses the per-frame latency of YuNet, Dlib HOG with and without upsampling, and RetinaFace at four input resolutions from 480p to 2160p. It examines how latency scales with frame area, separates fixed from pixel-proportional cost, quantifies the price of Dlib upsampling, evaluates detection quality per second of computation and derives admissible resolutions for frame-rate budgets of 1–30 fps. Between 1080p and 2160p latency grows almost in proportion to frame area; the resolution dependence of YuNet stems from frame resizing rather than inference; each Dlib upsampling step multiplies latency about 3.5-fold; and YuNet on a central processor together with RetinaFace on a graphics processor form the Pareto front. On this basis, practical design rules for edge deployment are proposed.
References
1. Guo J., Deng J., Lattas A., Zafeiriou S. Sample and computation redistribution for efficient face detection // International Conference on Learning Representations (ICLR 2022). – arXiv:2105.04714. – 2022.
2. Howard A.G., Zhu M., Chen B., Kalenichenko D., Wang W., Weyand T., Andreetto M., Adam H. MobileNets: efficient convolutional neural networks for mobile vision applications // arXiv preprint arXiv:1704.04861. – 2017.
3. Huang J., Rathod V., Sun C., Zhu M., Korattikara A., Fathi A., Fischer I., Wojna Z., Song Y., Guadarrama S., Murphy K. Speed/accuracy trade-offs for modern convolutional object detectors // Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017). – Honolulu, 2017. – P. 7310–7319.
4. Kortli Y., Jridi M., Al Falou A., Atri M. Face recognition systems: a survey // Sensors. – 2020. – Vol. 20, No. 2. – Art. 342. – DOI: 10.3390/s20020342.
5. Lin T.-Y., Dollár P., Girshick R., He K., Hariharan B., Belongie S. Feature pyramid networks for object detection // Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017). – Honolulu, 2017. – P. 2117–2125.
6. Sandler M., Howard A., Zhu M., Zhmoginov A., Chen L.-C. MobileNetV2: inverted residuals and linear bottlenecks // Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2018). – Salt Lake City, 2018. – P. 4510–4520.
7. Tan M., Pang R., Le Q.V. EfficientDet: scalable and efficient object detection // Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020). – Seattle, 2020. – P. 10781–10790.
8. Viola P., Jones M.J. Robust real-time face detection // International Journal of Computer Vision. – 2004. – Vol. 57, No. 2. – P. 137–154.
9. Wu W., Peng H., Yu S. YuNet: a tiny millisecond-level face detector // Machine Intelligence Research. – 2023. – Vol. 20, No. 5. – P. 656–665. – DOI: 10.1007/s11633-023-1423-y.
10. Zhang K., Zhang Z., Li Z., Qiao Y. Joint face detection and alignment using multitask cascaded convolutional networks // IEEE Signal Processing Letters. – 2016. – Vol. 23, No. 10. – P. 1499–1503.
This work is licensed under a 