Sacha Morin

Université de Montréal, Mila - Quebec Artificial Intelligence Institute

Papers

Total Citations

201

H-Index

About

Sacha Morin is a robotics researcher whose work sits at the intersection of 3D scene understanding, robot navigation, and large vision-language models. His most impactful contribution is **ConceptGraphs**, a seminal framework that constructs open-vocabulary 3D scene graphs from RGB-D data, enabling robots to perceive and plan in semantically rich, compact representations. This work has garnered over 190 citations since 2023, reflecting its significance in bridging perception and task-driven planning. Morin has also advanced monocular navigation, demonstrating that self-supervised Vision Transformers can learn coarse segmentation models with minimal annotated data—a practical breakthrough for deploying robots in environments like Duckietown. His **One-4-All** framework further tackles long-horizon navigation by introducing neural potential fields, offering a semi-parametric approach that sidesteps the pitfalls of end-to-end learning. Across these contributions, Morin consistently pushes toward representations that are both semantically expressive and computationally efficient, making his research highly relevant for embodied AI and real-world robotics. His work is especially notable for leveraging foundation models to reduce the need for task-specific training data, a direction that promises more adaptable and generalizable robotic systems.