An innovative way to mark attenence using face recognition.
Online education, though the need of the hour, has also given rise to a lot of instances of students missing classes and cheating. So, we need a robust and innovative system for attendance marking, which is altered for this particular situation.
The images are captured and saved during the registration of each student in a database. A model is trained to differentiate between real live fed video and photographs The model first predicts if it is a real human. If it is a real human face then the siamese network comes into action. The siamese network compares photo with all the photo using cosine similarity. If the similarity is above a certain threshold then the person is marked present and the date and time is also noted This data in turn will be saved as a csv file for future reference.
- The face-recognition package which has more than 99% accuracy detects face and encodes them
- This encoding is fed into a siamese network to compare the cosine similarity
- A threshold (95%) is set. Faces above this similarity are recognized
- A model is trained to classify real live image from photos which checks the liveliness
To run the project locally make sure to have the following dependencies installed on your system.
- flask
- flask-lgoin
- flask SQL Alchemy
- OpenCV
- pandas
- Keras
- Tensorflow
- face-recognition
- pyhton-3.x
- Clone the project
- Install all the dependencies
pip install -r requirements.txt
- Save photos in People folder
- execute while in the directory
python app.py
The model recognizes the person and marks his/her attendance while recognizing possible fakes like images.
