Transfer Learning for Children Face Recognition Accuracy

Abstract

Identifying the missed and kidnapped children at the later age will be a quit challenging process. To overcome the challenge, this research work has proposed a new Children Face Recognition (CFR) application using Artificial Intelligence (AI) system. To the best of our knowledge, the existence of children face image dataset, which is created in a fruitful process has not been reported in the earlier literature. Hence, to this consequence, this research work has addressed the problem of developing the children face recognition model with a suitable dataset. A model has been proposed by using machine learning pipeline that consists of pre-processing, feature extraction, dimensionality reduction and learning model. To this end, an attempt has also been made to classify the face images of children by training a multi-classification algorithm with the ensemble techniques such as Bagging and Boosting. During the dataset creation, 40,828 longitudinal face images of 271 young children of age from 4 to 14 years are captured over the duration of 30 months. This extensive experimentation has analyzed that few projection vectors of k-NN classifier has achieved a high accuracy of about 93.05%.