Detection of in-Car-Abandoned Children via Deep Learning

Abstract

Cases of children died in vehicle have been increased each year. Such incident sometimes may happen incidentally especially when children are seated at the rear seats and the problem occurs due to lacking of existing system in detecting children image in a car. Consequently, this study aims to detect the existence of "in-car-abandoned children" using deep learning algorithm. A set of children images model will be classified into two (2) classes; children and no-children via Convolutional Neural Network (CNN) classifier by integrating with programming language, namely TensorFlow. Interestingly, the proposed method can automatically learn pattern features and reduce the incompleteness caused by artificial design features. As a result, a model of sensor that can detect the whole children’s body in various poses with automatic tagging to the children’s image is designed. Accordingly, this study can assist to improve current vehicle systems and create awareness among parents regarding the importance of children’s safety.