Cognitive Analytics and Reinforcement Learning

5Virtual Makeup Try-On System Using Cognitive Learning

Divija Sanapala and J. Angel Arul Jothi*

Department of Computer Science, Birla Institute of Technology and Science Pilani, Dubai Campus, Dubai, UAE

Abstract

Virtual try-on technology helps consumers to examine how specific things look on them before purchasing. This makes the buying experience easier, excitement-filled, and mess free. Customers are less likely to return items because they can test them before purchasing. This paper implements a virtual makeup try-on system using deep learning. The proposed system comprises two parts. The first section employs a proprietary convolutional neural network (CNN) to extract face key points from the input image. The second part aims to extract the desired facial regions and then apply makeup products on it using the mask and wrap up methods. The proposed model is used to apply lipstick, eyebrows, and eyeliner. The CNN model employed consists of four convolution layers and five fully connected layers. Batch normalization, max pooling, and dropout layers are utilized between the CNN’s convolution layers. The Kaggle YouTube Faces Dataset is used in this work. For facial key point detection, the suggested model is compared to existing classical and deep learning algorithms. It is found that the proposed model has achieved an accuracy of 98.5%. It is user friendly and can help to increase market sales. It also helps users to test makeup virtually, saving time and money by avoiding ...