The Depth Estimation of 2D Content: A New Life for Paintings

The preservation, accessibility, and dissemination of historical artifacts to a wider audience have become increasingly important, and cultural institutions can achieve these goals through the digitization of cultural heritage. In recent years, artificial intelligence (AI) and machine learning (ML) techniques improve the virtualization of cultural artifacts for interactive experiences. In this work, we present a virtualization pipeline for the cultural heritage domain, focusing specifically on paintings, using AI techniques. We outline the basic workflow, including a thorough description of the comparison of various neural network models and their performance metrics. The proposed method creates an immersive experience for viewers to interact with paintings beyond observation. The approach utilizes 2.5D technology by applying depth maps of paintings using deep learning (DL) algorithms. The proof of concept was demonstrated on two real-life paintings of varying complexities, and this innovative approach holds potential for enhancing the appreciation and understanding of cultural heritage in museums and other cultural institutions.