What is my name?

Abstract

The thesis work introduces a new concept of predicting the first name from the facial image. Several studies show a relationship between first names and facial features. The prediction problem here is treated as an image classification problem. Here we have made a comparative analysis of various Ensemble learning techniques and a Transfer learning method. Dataset Name100 consists of 100 popular names of the U.S is used. The previous study was done on U.S. facename dataset. We have created an Indian facename dataset. The Indian dataset right now contains 5 Indian names. Thus we have used these two datasets for our experiments. The dataset mentioned here has no additional labelling cost apart from the name tags freely available on the internet. And we have also circulated a Google Form to collect the Indian data samples shown in this thesis. The prediction accuracy of this system is very low. Despite such a system�s low accuracy, the prediction was correct at more excellent rates than random chances. Though this system is imperfect for use in the real world today, it still has its applications in security and biometrics. And it also has the application of Face Name association. We have observed that misclassification happens amongst names of the same gender. While in the case of gender-neutral names, it is highly nondeterministic. Thus, a lot of research work needs to be done in this area.