"Computer Vision Meets Portrait Research"
by Prof Thomas Mandl, University of Hildesheim
Abstract
Digital Humanities research is focusing on enriching scholarship in Humanities and Cultural studies by employing digital methods for collecting, preserving and analysing artefacts. The paradigm of Distant Reading has proven to be especially productive. Since the Iconic Turn, research with images and visual material has itself established within the Humanities beyond the classic image sciences. For Digital Humanities, the development of appropriate tools and methods for Distant Viewing, which stands for the automatic analysis of large amounts of visual data with AI algorithms is still an emerging research field. In the last years, considerable progress has been made in image processing, especially through approaches of so-called Deep Learning. Thus, the classification of photographs is done based on algorithms, which not only learn the illustration but also, which aspects of the pictures need to be analysed. A prototypical system is a Convolutional Neural Network that combines many simple neurons as processors into complex architectures. This talk will briefly introduce CNNs. A review of approaches of Distant Viewing approaches and systems will be given. Then, the talk will report on experiences from working with image collections in two projects. One is about a collection of 32,000 early modern portraits. The other one deals with collections of pedagogical images mainly from children and youth literature. The goals include print type classification, object detection, similarity of publishers and face recognition on portraits. A discussion will introduce the challenges of processing historical data and working with concepts from the humanities.
Speaker’s bio
Thomas Mandl is a Professor of Information Science at the University of Hildesheim. He received his Doctorate in information science at the University of Hildesheim in 2000. He was appointed as an extraordinary professor at the same university in 2010. He is well known for his work in human-machine interaction (usability, method research, international aspects) and user-oriented evaluation in information retrieval. He is also the lead organiser in HASOC shared task – Hate Speech and Offensive Content Identification in English and Indo-Aryan Languages from 2019. His most recent work in applying computer vision to portrait graphics has created a new direction in digital humanities research.