Paul-Edouard Sarlin

Papers

Total Citations

190

H-Index

About

Paul-Edouard Sarlin is a computer vision researcher whose work centers on visual localization — the challenge of determining a camera's precise position and orientation within an environment. His research has made significant contributions to making localization systems more scalable, efficient, and practical for real-world deployment in robotics, autonomous driving, and augmented reality. Sarlin's most influential work, *OrienterNet* (2023, 78 citations), introduced a groundbreaking approach that enables neural networks to localize directly using widely available 2D public maps rather than expensive 3D point clouds, mirroring how humans naturally navigate using simple maps. This represents a fundamental rethinking of localization pipelines, dramatically reducing the computational and storage costs traditionally associated with the field. His earlier contributions established a strong foundation in hierarchical localization strategies. His 2019 paper on coarse-to-fine hierarchical localization (49 citations) and his paired 2018 works on leveraging deep visual descriptors (44 and 19 citations) demonstrated how combining learned image representations with structured search hierarchies could achieve robust, large-scale pose estimation even under challenging appearance changes. Collectively, Sarlin's research has helped shift visual localization from brittle, resource-intensive systems toward elegant, deployable solutions — earning him recognition as a leading voice in practical 3D scene understanding.