Alleviate Child Begging by Background-Aware Image Classification

Image classification models has various practical applications ranging from object recognition, facial recognition, sentimental analysis etc. In this work, we plan to utilize the capability of the existing image classification models to identify the child involved in begging, a kind of forced labour in many countries. In many developing nations children are tricked to begging due to circumstantial reasons or forced labour. The first step to address child begging is by identifying the children involved in begging, which currently requires a lot of manual tasks and coordination. Automation of identification of children involved in begging is the focus of the study. Various image classification models of Convolutional Neural Network (CNN) and variations of CNNs (VGG16, EfficientNet) are compared for the usage with the publicly available images for training the model. We propose the use of image background learning to improve the image classification models to detect child begging in public places with 99% accuracy.