Date(s) - 27/02/2020
12:00 pm - 2:45 pm
E-lab Room 0.16
Daniel van Strien (British Library) and Kaspar von Beelen (The Alan Turing Institute)
A growing challenge for digital historians is the diversification of their sources. Traditionally, these included tabular and textual data, but in recent years the analysis of visual collections has gained prominence as well. In this workshop, we introduce you to the basic principles of visual analytics and its application to humanities research. In particular, we show how to apply computer vision to historical maps. Maps serve as a rich resource for historical inquiry, but until very recently, they remained difficult to examine digitally–unless they came as georeferenced and manually vectorized data sets. We demonstrate, however, how computer vision now enables researchers to analyse spatial data in ways that are efficient, scaleable and considerably less resource-intensive.
We zoom in on specific elements of the computer vision pipeline such as transfer learning, image augmentation and other techniques, all which greatly facilitate the analysis of historical maps at scale. The hands-on part of the workshop will cover useful and practical methods for research and curation, including; classifying images automatically (including methods for automating some of the quality assurance steps of digitisation work), understanding maps-as-data, Working with large and small (<200) collections of images; evaluating a classification model and assess its performance: How well it is working? Or is it the wrong tool for the task at hand?
This tutorial will focus primarily on using Ordance Survey maps (digitized by the National Library of Scotland) but many of the techniques we cover can be adapted to other types of visual material such as photographs, art collections, or satellite imagery. The workshop is open to everyone: no coding skills are required, but please bring your own laptop!