Max Robotics — Global Service Robot Directory + US Certification

About

Daniel Yamins is a leading researcher at the intersection of cognitive science, computer vision, and embodied AI, whose work seeks to build machines that perceive and reason about the physical world like humans do. His key research areas include physical scene understanding, unsupervised object segmentation, and visually-guided task-and-motion planning. Yamins made a major contribution with his work on "Flexible Neural Representation for Physics Prediction" (91 citations), which introduced a hierarchical particle-based representation that allows neural networks to flexibly model complex physical dynamics—a foundational step toward human-like physical reasoning. He has also advanced unsupervised visual learning through his "Spelke Object Inference" framework, which leverages motion cues to segment objects in real-world images without labels, drawing on cognitive principles. In embodied AI, Yamins co-created the ThreeDWorld Transport Challenge (27+ citations), a benchmark that pushes the boundaries of physically realistic simulation for training agents in complex manipulation tasks. His recent work on "Counterfactual World Modeling" (2023) proposes a unified architecture for vision that could replace task-specific models, promising a paradigm shift in machine perception. With over 200 citations, Yamins’ research is shaping the future of AI systems that understand and interact with the physical world as seamlessly as humans do.