THOUGHT

The main aim of the project is to build a Face Recognition System. This project detects the face of a person in real time from stored database.Face Detection is the first step for Face Recognition which is followed by extracting features in form of embeddings from the input image. Test image is passed into the program and its features are calculated and these embeddings are compared with embeddings of our database and the person is recognized.

MODEL

Model for face recognition was trained for the following techniques:

• FaceNet
• Eigen Values
• Local Binary Pattern Histograms(LBPH)
• Convolutional Neural Networks(CNNs)

WORKFLOW

We developed a real time dataset by capturing a short video clip of the people precisely to be of 7-8 seconds. We further segmented it to fetch the frames by using a python code. It approximately resulted in 140-150 frames per person. Model is trained according to the above Flow charts.

RESULT

We reached to an accuracy of about 84.67% while working in real time with the model. We have been continuously analyzing the approach and making efforts to increase the accuracy of the system.