This project demonstrates a real-time face recognition system using the DeepFace library with live webcam input. It captures frames from a video stream, detects faces using MTCNN, generates face embeddings using ArcFace, and compares them to pre-saved embeddings to recognize individuals.
⚠️ Note: This implementation currently suffers from low FPS (frames per second), especially due to the high computation time of MTCNN and ArcFace on each frame in real time without GPU acceleration. It shold became faster under GPU acceleration
Screen.Recording.2025-07-11.111527.mp4
- Real-time face detection and recognition from webcam/IP camera
- Embedding generation using ArcFace
- Face detection using MTCNN
- Known vs Unknown classification using cosine similarity
- Modular structure for better scalability
- Custom embedding database (
embeddings.pkl) for known individuals
| Component | Technology |
|---|---|
| Face Detection | MTCNN (via DeepFace) |
| Face Recognition | ArcFace (via DeepFace) |
| Framework | Python 3.11 |
| Face Matching | Cosine Similarity |
| Embedding Storage | Pickle (.pkl) |
| Camera Input | OpenCV |
| Visualization | OpenCV GUI |
Libraries:
deepfaceopencv-pythonnumpypickleos,cv2,glob,datetime
Contributors:
Arin Ganguly(https://github.com/arin-ganguly)
Ankan Das(https://github.com/Ankandas2004)
Abhijit Dey(https://github.com/IamAbhijit2004)