Challenge-Response based protocol for human verification, which can be applied in e-KYC or any auto verification system.
The challenges include 2D head pose, random times of eye blinks, mouth opening (TODO), etc.
After WebRTC connection between server and client successfully formed, there are challenge commands appear in the webcam asking client to attempt to fulfill. After a certain number of challenges, the code will decide if the client pass the liveness test.
There are parameters to constrain or relax the verification challenge.
-
Python >= 3.10:
pip install -r requirements.txt- WebRTC - aiortc via streamlit-webrtc Framework for connection between client and server backend
-
MongoDB server, including pymongo (MongoDB Python Client)
The project mainly relied on facial processing techniques of MediaPipe (via its secondary library FaceAnalyzer) including:
- Face mesh detection (& drawing)
- Eye Blink, Mouth Opening & Head Pose calculations
- etc.
-
For simple test of challenge-response via direct Webcam, run face_anti_spoofing.py:
python face_anti_spoofing.py -
For Streamlit WebRTC server, run app.py:
streamlit run app.py