CILVR SEMINAR: Self-Supervised Learning- Going beyond images
In this talk, we will dive into self-supervised learning (SSL) and how to generalise ideas developed for images to other domains.
Click here to view the recording
In this talk, we will dive into self-supervised learning (SSL) and how to generalise ideas developed for images to other domains. We will start by looking into different SSL approaches while getting an understanding of the inner workings of popular methods. This will allow us to unify methods and show that most approaches are very similar.
With this knowledge, we will see that SSL can be applied on differential equations by leveraging domain-specific data augmentations, called Lie Point symmetries. Finally, we will discuss prediction based approaches (JEPAs) and see how they can be leveraged for video representation learning.