<![CDATA[A facial recognition-based security system for filing cabinets using ESP32-CAM and the face-api.js library is designed as a modern solution to enhance the security of storing important documents in office or organizational environments. This system utilizes facial recognition technology as a biometric authentication method, offering greater reliability compared to conventional systems. The ESP32-CAM serves as the device that captures users' facial images in real-time and transmits them to a web interface for further processing. Facial detection and feature extraction are performed using the face-api.js library, built on top of TensorFlow.js, leveraging deep learning techniques based on Convolutional Neural Networks (CNN). The models employed include TinyFaceDetector for face detection, faceLandmark68Net for determining facial landmark points, and faceRecognitionNet for generating a 128-dimensional face descriptor. Registered facial data is stored in a web-based database, allowing significantly greater storage capacity compared to local storage in the ESP32-CAM's memory. Verification is conducted by comparing the detected face descriptor against stored data using the FaceMatcher algorithm with a threshold of 0.6. Testing results indicate that the system can accurately recognize faces under adequate lighting conditions, although performance decreases under low-light intensity. This integration of hardware and software provides a more efficient and modern security solution.]]>
Copyrights © 2025
