Facial-Recognition

This repository provides the outline for creating a facial recognition model using a custom VGG CNN Model

Image Dataset

• The dataset should be based on the number (N) of individual faces one wishes to uniquely detect
• Each individual (person) would be an unique class and the output shall be a probability value for the input image against each class
• Please ensure that the number of images for each class (individual) is the same and is cropped to the face.
• There sould not be any other face in the images and the faces should be clearly visible
• The dataset is split 80-20 into Train and Test

Custom VGG Architecture

• Input Shape - 128 x 128
• Convolution Layer (32)
• Convolution Layer (32)
• Pooling Layer (2x2)
• Convolution Layer (64)
• Convolution Layer (64)
• Pooling Layer (2x2)
• Convolution Layer (128)
• Convolution Layer (128)
• Pooling (2x2)
• Flatten Layer
• FUC (128) Relu
• Output Later (Output - N) (Activation = Softmax; Loss = Categorical Crossentropy; Optimiser = Adam)

Using the Trained Model

• Use the file Model Generator_Facial Recognition_CustomVGG to train the model on the UKT Image dataset and save the model file (.h5) at a desired location
• Use the file Video_Facial Recognition_CustomVGG_CNN by inserting the Model file name and location in order to predict gender with the computer Webcam