Deep Learning And SVM Based Missing Child Identification System
K.R.Thukaram Rao, Raj Purohith Pooja Kumari, Puli Sravanthi, B.Rakshitha, Tasab Yaseen Vali, G Sai Mythily
In India countless children go missing every year. Of the missing children cases, most of the children
go untraced. The current paper shows a new application of deep learning methodology to recognize the
reported missing child from the images of numerous children present, using face recognition. The public
can submit pictures of suspected child into a shared portal with landmarks and comments. The image
will be automatically matched with the enrolled images of missing child from the repository. Input child
image is classified and picture with highest match will be chosen from the missing children database.
For this purpose, a deep learning model is learnt to appropriately identify the missing child from the
missing child image database given, based on the facial image posted by the public.
Convolutional Neural Network (CNN), a very powerful deep learning method for image based
applications is employed here for face identification. Face descriptors are obtained from the images
based on a pretrained CNN model VGG-Face deep architecture. Unlike regular deep learning usage,
our algorithm utilizes convolution network as a high level feature extractor and the child recognition is
implemented by the SVM classifier trained. Selecting the best performing CNN face recognition model
as VGG- Face and properly training it gives a deep learning model that is invariant to illumination,
noise, contrast, occlusion, child's age, and image pose and it outperforms the previous methods on face
recognition based missing child detection.
