DoIKnowYou (DIKY) is an implementation of a Face Recognition framework using Haar Cascade classifiers inspired from the CVPR 2001 paper by Viola-Jones that can handle multiple client verification requests on a server using Multithreading and Synchronisation Policies.
The following libraries are needed:
- OpenCV 4.4.0
On Ubuntu 20.04, you can install with: pip install opencv-contrib-python
Please check dependencies and proceed. For maximal performance you can build from source with CUDA support.
-
Python 3.8.3
-
Python packages (newer packages will likely work, though these are the exact versions that we used):
numpy>=1.17.3
socket
tqdm
threading
Setting up a virtual environment like virtualenv will help keep your Python environment safe. We recommend installing all dependencies using this.
There are two important files, Client.py and Thread_Server.py and their dependency files. Fork and clone this repository in your local machine. Open one terminal to run the Thread_Server.py and open multiple terminals for each instance of Client.py using:
python3 client.pypython3 server.pyIn the client side, a Enter your name prompt will come and upon entering, a virtual terminal will be created. The server side will bind to localhost:5000 after launching. If the port is busy, please change the port address in both client and server side.
- If a log comes,
[RDY] Socket is now Deployed, we are now ready to listen for client requests. Else restart the program.
To listen for client requests, use
listenTo register a new user using a video, use
trainVideoA prompt will appear asking for video location, enter the absolute/relative address of the video. A consequent prompt will appear, asking to annotate the video with the name of the person.
To register a new user using a webcam, use
trainWebcA prompt will appear, asking to annotate the video with the name of the person.
To quit, use
quitTo send a verification request using a video, use
videoA prompt will appear asking for video location, enter the absolute/relative address of the video. A consequent prompt will appear, asking to annotate the video with the name of the person. After this a window will open up, showing the video and the detected person.
To register a new user using a webcam, use
webcamTo quit, use
quitTo avoid unexpected killed threads and orphaned processes, do not use this command in between client requests. System may experience a lag and you might have to hard-reboot. You have been warned.
After login on Server:
After login on a Client:
A 4 client request session to showcase Multithreading:
We have trained face recognition on two videos of actors Will Smith and Emma Watson, because we are in The Pursuit of Happyness and Hermione Granger is 💕 The videos are provided in the videos folder for retraining. Feel free to add more and go wild.
Note: You need IEEE Access for accessing these papers. We strongly discourage pirated websites. Please support the research community.
- P. Viola and M. Jones, "Rapid object detection using a boosted cascade of simple features," Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001, Kauai, HI, USA, 2001, pp. I-I, doi: 10.1109/CVPR.2001.990517.
- RGB-H-CbCr Skin Colour Model for Human Face Detection, Nusirwan Anwar bin Abdul Rahman et al., Faculty of Information Technology, Multimedia University.
- L. Cuimei, Q. Zhiliang, J. Nan and W. Jianhua, "Human face detection algorithm via Haar cascade classifier combined with three additional classifiers," 2017 13th IEEE International Conference on Electronic Measurement & Instruments (ICEMI), Yangzhou, 2017, pp. 483-487, doi: 10.1109/ICEMI.2017.8265863.