Bernhard Zeisl

Papers

Total Citations

H-Index

About

Bernhard Zeisl is a computer vision and robotics researcher whose work centers on visual localization, sensor calibration, and the development of scalable systems for real-world deployment in robotics and Augmented Reality (AR). His research addresses some of the most demanding challenges in these fields, including operating in large-scale, repetitive environments where accurate pose estimation is notoriously difficult. Zeisl's most impactful contribution, "Efficient Descriptor Learning for Large Scale Localization" (2017, 27 citations), demonstrates his focus on making keypoint-based visual mapping both computationally efficient and robust — a critical requirement for AR and autonomous systems with limited processing resources. His work on RGB-D sensor auto-calibration (2016, 11 citations) further reflects his commitment to practical, deployable solutions, enabling better fusion of depth and image data without relying on artificial calibration targets. More recently, his revisitation of visual-inertial localization at scale (2020) underscores his sustained interest in pushing localization systems toward greater speed, robustness, and real-world applicability. Across his research, Zeisl consistently bridges the gap between theoretical innovation and practical system deployment, making meaningful contributions to how machines perceive and navigate the physical world.