Feature Extraction of Image Content by AI and Its Application

Abstract

Recent rapid developments of image AI technologies are expanding the possibilities for the use of image materials in digital archives. The article describes how two representative image AI technologies, Convolutional Neural Network (CNN) and Vision Transformer (ViT), can be used for image recognition and feature extraction, and shows classification of similar images as an example. As an application example, an image archive exploration support Web application is described in detail. The application combines image placement based on image similarity and metadata similarity using natural language AI techniques to effectively explore visually and thematically relevant material.