Yusong Li

Papers

Total Citations

H-Index

About

Yusong Li is a leading researcher in computer vision and spatial intelligence, specializing in end-to-end visual localization for dynamic indoor environments. Their work bridges the gap between 2D image understanding and 3D pose estimation, addressing the critical challenge of achieving centimeter-level accuracy where traditional sensor-based methods (Bluetooth, Wi-Fi) fall short. Li's most cited paper, "TransCNNLoc" (2023, 9 citations), introduced a pioneering pixel-level learning framework that enables robust 2D-to-3D pose estimation in cluttered, changing indoor scenes—a breakthrough for augmented reality and robotics. Building on this, their 2025 work on BIM-based indoor navigation (4 citations) integrates visual localization with ARCore, demonstrating a seamless pipeline for real-time AR navigation that leverages building information models. This research directly tackles the limitations of conventional localization, offering a scalable solution for fine-grained applications like autonomous robot positioning and immersive AR experiences. Li's contributions are shaping the future of indoor spatial computing, providing the foundational technology for next-generation navigation systems that operate reliably in the unpredictable, dynamic spaces where people live and work.