Yoni Friedman
Papers
2
Total Citations
20
H-Index
2
About
Yoni Friedman is a computer vision researcher whose work bridges artificial intelligence and cognitive science to tackle one of the field’s most persistent challenges: unsupervised, category-agnostic image segmentation. His research centers on developing self-supervised learning methods that enable machines to parse real-world images into meaningful objects without requiring labeled training data. Friedman’s key contribution lies in operationalizing the cognitive concept of a “Spelke Object”—a coherent physical entity that moves as a unit—to learn static grouping priors from motion self-supervision. This innovative approach, detailed in his highly cited 2022 paper, demonstrates how temporal cues in video data can teach models to recognize object boundaries in static images, effectively bootstrapping segmentation from how things move in the world. With over 20 citations to date, this work has quickly gained traction for its elegant solution to a notoriously difficult open problem. By grounding computer vision in developmental psychology principles, Friedman is helping to create more robust, human-like visual perception systems that can understand scenes without exhaustive manual annotation.
Research Focus
Key Achievements
Top Papers
- 1Unsupervised Segmentation in Real-World Images via Spelke Object Inference17 citations · 2022
- 2