SYSTEMATIC ANALYSIS OF A DEEP-LEARNING-BASED REAL-TIME FACE RECOGNITION SYSTEM
Received: 2026-05-29
Published: 2026-06-06
Abstract
This paper provides a detailed account of the full architecture and the operating mechanism of each stage of a deep-learning-based modular system for real-time person identification (face recognition) in surveillance camera streams. The system consists of a five-stage pipeline: frame acquisition, face detection (RetinaFace/MTCNN), alignment, embedding (FaceNet/ArcFace), and comparison via cosine similarity. Embedding vectors are searched rapidly through a PostgreSQL + pgvector store. In the experimental part, the comparison logic was evaluated on a synthetic embedding set of 50 identities, yielding AUC = 1.000, an Equal Error Rate (EER) of 0%, an accuracy of 1.000 at a 0.6 threshold, and an F1-score of 0.995. The paper explains each component at a “why and how” level so that a technical specialist can fully understand and reproduce the system.
Keywords
List of references
-
Schroff F., Kalenichenko D., Philbin J. FaceNet: A unified embedding for face recognition and clustering // IEEE CVPR. 2015. P. 815–823.
-
Deng J., Guo J., Yang J. et al. ArcFace: Additive Angular Margin Loss for Deep Face Recognition // IEEE TPAMI. 2022. Vol. 44, No. 10. P. 5962–5979.
-
Deng J., Guo J., Ververas E. et al. RetinaFace: Single-shot multi-level face localisation in the wild // IEEE/CVF CVPR. 2020. P. 5202–5211.
-
Zhang K., Zhang Z., Li Z., Qiao Y. Joint face detection and alignment using multitask cascaded convolutional networks (MTCNN) // IEEE Signal Processing Letters. 2016. Vol. 23, No. 10. P. 1499–1503.
-
Abid M.M., Mahmood T., Ashraf R. et al. Computationally intelligent real-time security surveillance system in the education sector using deep learning // PLOS ONE. 2024. Vol. 19, No. 7. e0301908.
-
Firmansyah A., Kusumasari T.F., Alam E.N. Comparison of face recognition accuracy of ArcFace, FaceNet and FaceNet512 models on DeepFace framework // ICCoSITE. 2023. P. 535–539.
-
Iype J.K., Sebastian S. Comparative analysis of state-of-the-art face recognition models: FaceNet, ArcFace, and OpenFace // ICCSST 2023, CCIS, vol. 1973. Springer, 2024.
-
Optimizing face recognition inference with a collaborative edge–cloud network // Sensors (MDPI). 2022. Vol. 22, No. 21. Art. 8371.
About the Authors
License
This work is licensed under a Creative Commons Attribution 4.0 International License.