Computer Vision for the Humanities and Social Sciences: An Introduction to Deep Learning for Image Classification

Masterclass Description

Over the last 10 years, the field of computer vision, which seeks to gain a high-level understanding of images using computational techniques, has seen rapid innovation. For example, computer vision models are able to locate and identify people, animals and thousands of objects on images with high levels of accuracy. This technological innovation promises the same innovation for images that the combination of Optical Character Recognition/NLP (Natural language processing) techniques caused for texts. They open up a part of the digital archive for large-scale analysis, which, until now, has been left uncovered: the millions of images in digitized books, newspapers, periodicals, and historical documents. Recent applications of these methods for humanities-related questions include projects that extract images from historical newspapers, analyse ‘Visual Style in Two Network Era Sitcoms’, and work with large digitised map collections.

By the end of the workshop you should be in a position to identify opportunities to use computer vision in your discipline, know how to identify candidate datasets and have an appreciation of what’s involved in a project involving computer vision. The hands-on aspects of the workshop will give you exposure to key considerations when working with deep learning and you’ll be able to tweak provided examples conduct your own prototype experiments simply in a lab environment.

The workshop will:

• Provide an introduction to deep learning-based computer vision methods for humanities research.

• Give an overview of the steps involved in training a deep learning model and possible approaches using code and no code solutions.

• Discuss some of the specific considerations around using deep learning/computer vision for humanities research.

• Help you decide whether deep learning might be a useful tool for you.

Suggested Pre-requisites

You are welcome to join without the experience below, but having some familiarity with these topics will be helpful. If you don’t have this experience, you can still participate in the workshop's hands-on sections.

• Some familiarity with Python or another programming language will be helpful. Specifically, it would be beneficial to understand how to use variables, indexing and to have some familiarity with using methods from external libraries. You are still welcome to join without this existing experience.

• Some basic familiarity with using Jupyter Notebooks will be an advantage, i.e. knowing how to run the code included in a Jupyter Notebook. If you are unfamiliar with notebooks, you may find the introduction to Jupyter Notebooks Programming Historian lesson a helpful resource prior to the session.

Technical Requirements: participants will need a laptop or personal computer with internet access and a modern browser (Firefox or Chrome preferred). It might be possible to follow materials with a tablet, but I won’t be able to troubleshoot issues with this setup.

Speaker Biography

Daniel van Strien is a Machine Learning Librarian at Hugging Face a company focused on democratizing good machine learning. At Hugging Face Daniel focuses on helping improve the Hugging Face Hub, a central repository for sharing machine learning models, datasets and demos. Prior to this Daniel worked as a Digital Curator at the British Library working on the Living with Machines project.

Event Information  

This event will take place in the Digital Scholarship Centre on the 6th floor of the Main Library. Please inform us of any access requirements by emailing cdcs@ed.ac.uk. Further details about how CDCS uses your information obtained from booking onto our events can be found in our Events Privacy Statement. 

As of March 2022, the government formally removed all Covid restrictions in the UK. We ask that you continue to be considerate of others’ personal space, and please do not attend if you feel unwell or have any of Covid symptoms.