Segun Aina; Mosunmola Oluwabusola Adeniji; Aderonke Rasheedat Lawal; Adeniran Isola Oluwaranti

The human face is considered to be the seat of man’s identity and information such as age and ethnicity are often automatically deduced from the face by people. However, deducing the same information by a computing system is not a straight forward process and have in recent years be powered by Convolutional Neural Networks (CNN). CNN can automatically extract hidden patterns in data. These hidden patterns are often complex to represent using hand-crafted representation methods. Although automated classification of demographic traits such as age, gender and ethnicity is a well-studied research problem, it is still far from being considered a solved problem for Nigerian ethnic groups. In this paper, a CNN model for ethnicity classification of Nigerians from facial images is proposed based on transfer learning techniques conducted on VGG-16 architecture. The model is evaluated on a dataset consisting of facial images of Yoruba, Hausa and Igbo ethnic groups of Nigeria. The developed VGG-16 based ethnicity classification model had an overall accuracy of 92.86%, with the precision, sensitivity and specificity shedding more light on the model’s performance