<ul><li>Object-centric learning (OCL) focuses on learning representations that encode isolated objects in a scene.</li><li>Recent advancements in sample-efficient segmentation models allow the separation of objects in the pixel space.</li><li>A new training-free probe called Object-Centric Classification with Applied Masks (OCCAM) outperforms slot-based OCL methods in out-of-distribution (OOD) generalization.</li><li>Challenges in real-world applications and fundamental questions related to object perception in human cognition remain.</li></ul>

Are We Done with Object-Centric Learning?

Discover more