Age Estimation using Lightweight Convolution Neural Network

Introduction: Currently, age estimation is significant in sectors such as age-specific human-computer interaction, security control, surveillance monitoring, Electronic customer relationship management, and forensic art. Nowadays makeup is used by all groups of people. Makeup is now commonly used to boost an individual's confidence and self-esteem. Makeup can give a person a younger or older appearance depending on their makeup choice. This transformation is caused due to varied makeup products such as eyeliner, lipstick, foundation, compact powder, mascara, and eye pencil. These products impact the real face of the individual by concealing the age spots, scars, pimples, and wrinkles, highlighting the cheeks, and brightening the facial parts. Due to the transformation in the look, the apparent age varies from the real age. The estimation of real age is essential in various real-time applications. Objectives: The application of makeup is a great challenge in estimating the age of the person in vision-based technology. This manuscript deals with deep learning-based age estimation with makeup. Methods: The proposed framework is composed of data augmentation, face detection, face alignment, and age estimation. The data augmentation techniques used in our methodology are average blur, gaussian blur, histogram equalization, color jittering, bilateral filtering, detexturization, unsharp filtering, and gamma contrast. Multi-task Cascaded Convolutional Network (MTCNN) is adopted for face detection and alignment. Light CNN is used for age estimation with makeup. Results: The methodology attains an MAE of 5.58 on the self-built Facial Makeup for Male and Female (FMMF) database. Conclusions: This manuscript presents a deep learning based-age estimation with facial make-up and its results are also evaluated. The developed mythologies will improve the quality of applications such as human-computer interaction, surveillance systems, commercial development, content-based indexing, and searching, demographic studies systems, targeted advertising, and biometric system.