Week 6: Self-supervised learning and imaging foundation models

Slides 6-1

Papers for Thursday

He et al.: MAE
Chen et al.: UNI

Animation clips

Each clip comes in two versions. The narrated version has an AI-generated voice and captions; turn them on with the CC button. The silent version is the same animation with its on-screen captions only.

Masked reconstruction: learning by filling in gaps

1 min 35 s

Silent version Transcript Captions (.vtt)

Contrastive learning: two views, one embedding

1 min 40 s

Silent version Transcript Captions (.vtt)

Linear probing: what did the encoder learn?

1 min 25 s

Silent version Transcript Captions (.vtt)

Everything for this week

The Week 6 folder in the repository ↗, with the animation sources, the slides and the session guide.