Face Recognition to Determine Visitor Attraction Using Residual Deep Neural Network

The development of social media in the form of websites and android applications should be appreciated. The background of this research is the stage to develop the software needed to know the age range. The objective thing to observe is through the visitor’s face on the web or application. The purpose is to avoid false information and get the proper process flow. The Technology used is deep learning for facial image recognition. The methods use Convolutional Neural Network with residuals because the advantage is using multi-branch layers but having a stable training process. The use of augmentation is needed to increase the variety of facial recognition image positions and to overcome unbalanced classes. The dataset used consists of seven categories, namely Children, Youth, Early Workers, Middle Ages, Pre-Retirement, Retirement, and Old People, which Bappenas regulate. Each folder contains about 1157 images with a total data of 8,105 images. The training process results can obtain a model with an average accuracy of 99.08% and a computational training time of about three hundred minutes. The model’s accuracy is 99.08%, with MSE 0.0037, RMSE 0.0610, and MAE 0.0037. Testing time is about 2 seconds.