Modeling Expectation Violation in Intuitive Physics with Coarse Probabilistic Object Representations
- Kevin A. Smith, Lingjie Mei, T. Ullman
- PhysicsNeural Information Processing Systems
- 2019
ADEPT, a model that uses a coarse (approximate geometry) object-centric representation for dynamic 3D scene understanding, outperforms standard network architectures in discriminating physically implausible scenes, and often performs this discrimination at the same level as people.
The fine structure of surprise in intuitive physics: when, why, and how much?
- Kevin A. Smith, Lingjie Mei, T. Ullman
- PhysicsAnnual Meeting of the Cognitive Science Society
- 2020
It is shown that the timing and degree of surprise can be explained by an object-based model of intuitive physics, and three distinct measures are asked: adults are asked to judge how surprising a scene is, when that scene is surprising, and why it is surprising.
FALCON: Fast Visual Concept Learning by Integrating Images, Linguistic descriptions, and Conceptual Relations
- Lingjie Mei, Jiayuan Mao, Ziqi Wang, Chuang Gan, J. Tenenbaum
- Computer ScienceInternational Conference on Learning…
- 30 March 2022
We present a meta-learning framework for learning new visual concepts quickly, from just one or a few examples, guided by multiple naturally occurring data streams: simultaneously looking at images,…
FALCON: F AST V ISUAL C ONCEPT L EARNING BY I N TEGRATING I MAGES , L INGUISTIC DESCRIPTIONS , AND C ONCEPTUAL R ELATIONS
- Conceptual Relations, Lingjie Mei, Jiayuan Mao, Ziqi Wang
- Computer Science
- 2022
We present a meta-learning framework for learning new visual concepts quickly, from just one or a few examples, guided by multiple naturally occurring data streams: simultaneously looking at images,…