Hybrid Deep Learning and Multi-Class SVM Approach for Missing Child Identification

Abstract

In India a countless number of children are reported missing every year. Among the missing
child cases a large percentage of children remain untraced. This paper presents a novel use
of deep learning methodology for identifying the reported missing child from the photos of
multitude of children available, with the help of face recognition. The public can upload
photographs of suspicious child into a common portal with landmarks and remarks. The
photo will be automatically compared with the registered photos of the missing child from
the repository. Classification of the input child image is performed and photo with best
match will be selected from the database of missing children. For this, a deep learning
model is trained to correctly identify the missing child from the missing child image
database provided, using the facial image uploaded by the public. The Convolutional Neural
Network (CNN), a highly effective deep learning technique for image based applications is
adopted here for face recognition. Face descriptors are extracted from the images using a
pre-trained CNN model VGG-Face deep architecture. Compared with normal deep learning
applications, our algorithm uses convolution network only as a high level feature extractor
and the child recognition is done by the trained SVM classifier. Choosing the best
performing CNN model for face recognition, VGG-Face and proper training of it results in a
deep learning model invariant to noise, illumination, contrast, occlusion, image pose and
age of the child and it outperforms earlier methods in face recognition based missing child
identification. The classification performance achieved for child identification system is
99.41%. It was evaluated on 43 Child cases