Artificial Intelligence for Hair Color Rendering

The importance of hair coloring has grown as a trend and people who want to change their hair color often look for recommendations at hair salons without knowing the final result. The purpose of a hair color rendering application is to provide a digital tool that offers color recommendations by applying hair colorimetry and allowing virtual preview of a new shade. The application collects data on eye, skin and hair color in a MySQL database, uses the Support Vector Machine algorithm and the DeepLabv3 model to predict color and segment hair in images. The evaluation results of the algorithm show a value of 0.78 in training and testing accuracy, manifesting adequate ability to predict color. For the evaluation of the model, solid quality metrics are used such as: recall 98%, Dice Coefficient 98% and IoU 97%, which have an adequate capacity to segment hair. In summary, this application offers personalized and realistic advice, combining creativity with technology to provide a favorable change in the user’s appearance.