Convolutional Neural Network Classification On 2d Craniofacial Images

Abstract

This study presents a craniofacial superimposition consisting of filtering, feature extraction, classification
for skull modeling. Initially, the Gaussian filter removes the noise and Otsu thresholding extracts the
features. The extracted features is sent as a trained data to the Convolutional Neural Network (CNN).
The experiment is carried out on 100 input image set that encompasses both the cranial and facial image.
The experiment is conducted on new input image and it is then applied directly to the CNN classifier. The
experimental results show that the CNN classifier achieves higher classification rate for 2D landmark
and skull modelling than the existing methods.