College of Engineering

Department of Computer Engineering  / College of Engineering at University of Mosul discussed a master’s thesis on “Using Machine Learning for Person’s Classification Based on Images of Hand” for the postgraduate student (Sarah Ibrahim Fathi) on Thursday, March 07, 2024.

This thesis investigates a novel method of using hand photographs for kinship detection. The first step in this work was the creation of a new dataset, which is named the Mosul Kinship Hand (MKH) because the available hand image datasets do not contain kinship ground truth. This dataset was collected using a mobile phone camera. They consisted of 634 images for 81 individuals from 14 families.

This research also presents the use of this dataset in kinship prediction using machine learning. Both handcraft and deep transfer learning were used in feature extraction from hand images in ten different experimental scenarios. Neural network classifiers were designed and trained to predict the membership of any person in one of the families in the MKH dataset.

As a promising way to use the hand as a kinship indicator that may enhance the use of the earlier methods, the results of this novel approach showed that the hand possesses biometric characteristics that may be used for familial classification with an essential accuracy range of 75% to 93%.