Verification of Musical Education with Face Images Using Convolutional Neural Network and Vision Transformers

This research investigates whether machine learning techniques can be used to study the relationship between physical appearance and personality or other characteristics. Specifically, it explores the hypothesis that a person's profession may influence their facial features. The study focuses on musicians of a symphony orchestras as a specific example. A dataset of photographs was collected, including both symphony orchestra musicians and non-musicians. Three neural networks were used to solve the problem of classifying the images into two categories (musicians vs. non-musicians): VGG-Face, CLIP-ViT, and Google ViT. These neural networks extracted numerical representations (embeddings) from the images. The embeddings were then used to train support vector machines, a type of machine learning algorithm. The study compares the amount of information captured by the embeddings from each of the three neural network models.