Paper Key : IRJ************019
Author: Allagadda Venkata Sagarika
Date Published: 04 Mar 2024
Abstract
In India, a significant number of children are reported missing each year, with many cases remaining unresolved. This paper introduces an innovative approach that leverages deep learning techniques for identifying missing children from a large pool of available photographs through face recognition. The public is encouraged to upload images of suspicious children to a centralized platform, where these photos are automatically compared against registered photos of missing children. Classification is then performed to determine the best match from the database of missing children. To achieve this, a deep learning model is trained specifically for identifying missing children using facial images uploaded by the public. The Convolutional Neural Network (CNN), a powerful technique for image-based applications, is utilized for face recognition. Face descriptors are extracted using a pre-trained CNN model, VGG-Face deep architecture. Unlike traditional deep learning approaches, our algorithm focuses on using the convolutional network as a feature extractor, while child recognition is handled by a trained SVM classifier. By selecting the best-performing CNN model, VGG-Face, and ensuring proper training, our deep learning model becomes robust to various factors such as noise, illumination, contrast, occlusion, image pose, and the child's age, outperforming previous methods in missing child identification based on facial recognition.