Machine Learning Street Talk (MLST) cover image

#55 Self-Supervised Vision Models (Dr. Ishan Misra - FAIR).

Machine Learning Street Talk (MLST)

00:00

Navigating Self-Supervised Learning in Vision Models

This chapter explores the intricacies of self-supervised learning in computer vision, focusing on the transformation of images into vector representations and the critical role of data augmentation. It contrasts self-supervised methods with traditional supervised techniques, addressing challenges such as mode collapse and the implications of human knowledge in model training. The discussion highlights recent advancements in algorithms and their capacity to improve performance in various visual tasks without relying heavily on labeled data.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app