N. Siddharth

Papers

Total Citations

H-Index

About

N. Siddharth is a researcher whose work sits at the intersection of computer vision, robotics, and cognitive science, with a particular focus on how machines can learn to perceive and interact with the physical world. His early, highly-cited paper "Learning physically-instantiated game play through visual observation" (23 citations) pioneered an integrated vision and robotic system capable of learning to play board games like TIC TAC TOE and HEXA-PAWN through direct visual observation, using custom hardware designed specifically for this learning task. This work demonstrated a novel approach to grounding abstract game rules in physical interaction. Expanding on this theme, Siddharth introduced a "visual language model" for estimating object pose and structure, drawing a compelling analogy between the generative grammar of human language and the compositional nature of physical objects. His work "Seeing Unseeability to See the Unseeable" tackled the challenging problem of reasoning about occluded structures, developing a framework that uses visible evidence and consistency models to infer hidden portions of objects. Through these contributions, Siddharth has advanced our understanding of how machines can bridge the gap between visual perception and physical reasoning.