Find the workshop material on GitHub: https://github.com/i4Ds/AMLD-2022-Visual-Disinformation

With the advent of the internet, mobile devices and social media, we are currently experiencing an unprecedented democratization concerning the publication of information. Unfortunately, information is shared not always with good intent in mind---fake news and hate speech are on the rise.

Internet memes have become a powerful tool to shape online narratives. They are used for nefarious purposes by different actors around the globe as diverse as e.g. the alt-right in the USA or the Taliban in Afghanistan. Image-with-Text (IWT) memes offer a unique multimodal combination of image and text data, which can be analyzed using state-of-the-art deep learning methods from computer vision and natural language processing (NLP).

In this workshop, the attendees will be introduced to the topic of visual disinformation and the role internet memes play in it. They will learn how to use transfer learning to adapt pre-trained models from both computer vision and NLP to analyze IWT memes . We will be using Python and deep learning modules from fast.ai and PyTorch libraries to work on multimodal data.

At the end of the workshop, the attendees will be familiar with the topic of visual disinformation and different IWT meme datasets. They will know how to utilize pre-trained models from computer vision and NLP to perform different tasks (e.g. image classification) and characterize visual content (e.g. emotionality of an image). Finally, attendees will have explored multimodal deep learning and visual disinformation as potential areas of research.

• Basic programming knowledge with Python
• Basic knowledge of Machine Learning/Deep Learning
• Interest in geopolitics and social cybersecurity
• Your own laptop
• No previous knowledge in computer vision and NLP required
• Workshop repo: https://github.com/i4Ds/AMLD-2022-Visual-Disinformation