Repo for the Image as data tutorial at the IC2S2 23 conference.
The material covers the full pipeline of how to use images as data - from setting up an envrionvent and installing packages to creating deep learning models. The material is created in python, using primarely Pytorch. The materials consists of jupyter notebooks, with links to readings and instructional videos.
Material on Colab: https://colab.research.google.com/drive/1WS9cJmRuME6Z2o_rIK06rWQYZd2XkpX0#scrollTo=ZV5kg9hVIr_U
https://colab.research.google.com/drive/1LEiz3C2NUoXVYnLwnA-YbnXgHQPhe9zR#scrollTo=ZV5kg9hVIr_U (solution)
All materials are also available here, but have not been tested locally so you may need to adapt. Feel free to reuse, alter, distribute etc. with proper atribution.
The tutorial is meant to be compleated in 3-4 hours, in a "flipped classroom" style, meaning students go through the notebook themselves, reading the material, whaching the videos and completing the excercises - with help from instructors. A notebook with solutions is also availbe.
The material has been developed by August Lohse, Center for Social Data Science, University of Copenhagen Simon Polichinel von der Maase, The Peace Research Institute Olso (PRIO)