Oisin Mac Aodha

Papers

Total Citations

H-Index

About

Oisin Mac Aodha is a leading researcher at the intersection of computer vision, robotics, and augmented reality (AR), with a focus on enabling machines to perceive and interact with 3D scenes in a generalizable, open-vocabulary manner. His work addresses fundamental challenges in scene understanding and precise manipulation, bridging the gap between visual perception and physical action. Mac Aodha’s most cited paper, "Click to Grasp: Zero-Shot Precise Manipulation via Visual Diffusion Descriptors" (2024, 7 citations), introduces a novel approach that leverages diffusion models to achieve precise, generalizable robotic grasping without task-specific training—a breakthrough for handling objects with ambiguous visual or geometric parts. In "Heightfields for Efficient Scene Reconstruction for AR" (2023, 5 citations), he proposes a real-time, learning-based method that reconstructs 3D scenes from RGB images, outperforming traditional depth-fusion techniques for augmented reality applications. His work "AirPlanes: Accurate Plane Estimation via 3D-Consistent Embeddings" (2024, 3 citations) demonstrates that combining clustering with 3D-consistent features yields a surprisingly strong baseline for planar surface extraction, critical for robotics and AR. Additionally, "VL-Fields: Towards Language-Grounded Neural Implicit Spatial Representations" (2023, 3 citations) introduces a neural implicit representation that fuses geometry with vision-language features, enabling open-vocabulary semantic queries. With a consistent focus on zero-shot generalization and efficient 3D understanding, Mac Aodha’s research is shaping the future of intelligent, interactive systems.