• Dev Sharma

:Face recognition plays a vital role in security, authentication, and surveillance applications. This research proposes an optimized face recognition system that captures images based on an experimentally determined Eye Aspect Ratio (EAR) threshold of 0.23, ensuring high-quality image acquisition. The captured images are encoded in base64 format and temporarily stored before being processed using VGG-Face for feature extraction. If the extracted feature vector does not match any stored vectors within a cosine similarity threshold of 0.8, the new embedding is saved in a text file in base64 format for future reference. This method reduces storage requirements while ensuring efficient retrieval and security. Experimental results show that this approach enhances recognition accuracy, minimizes false positives, and optimizes computational efficiency. The system achieves a 96.7% measured accuracy, comparable to leading face recognition models such as Facenet512 (98.4%) and ArcFace (96.7%), outperforming models like OpenFace (78.7%) and DeepFace (69.0%). This study contributes to real-time authentication systems by improving image quality during capture, implementing an effective storage mechanism, and maintaining a high recognition accuracy. The results indicate that optimizing EAR-based image selection and efficient encoding techniques significantly improve the reliability of face recognition systems.

Face Recognition Eye Aspect Ratio (EAR) Base64 Encoding Image Quality Enhancement Feature Extraction VGG-Face Cosine Similarity Storage Optimization Real-time Authentication Face Embedding MTCNN (Multi-task Cascaded Convolutional Networks) Facenet512 ArcFace Deep Learning Biometric Authentication Surveillance Systems Computational Efficiency Real-world Accuracy Adversarial Attack Resistance DeepFace and DeepID Comparison

" Optimized Face Recognition System Using Eye-to-Ear Ratio and Base64 Encoding for Efficient Storage", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 4, page no.l268-l272, April-2025, Available :http://www.jetir.org/papers/JETIR2504B31.pdf

" Optimized Face Recognition System Using Eye-to-Ear Ratio and Base64 Encoding for Efficient Storage", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 4, page no. ppl268-l272, April-2025, Available at : http://www.jetir.org/papers/JETIR2504B31.pdf