shubham0204/FaceRecognition_With_FaceNet_Android

Embeds FaceNet's 128-dimensional face vectors into a TensorFlow Lite model for real-time classification without retraining—users simply organize reference images in device folders by subject name. Combines MLKit for face detection bounding boxes with CameraX frame analysis and Kotlin Coroutines for asynchronous processing, computing L2-norm or cosine similarity scores across multiple reference images per subject to determine best matches.

329 stars. No commits in the last 6 months.