Honghua Li

Alibaba Group (China), Shandong University

Papers

Total Citations

135

H-Index

About

Honghua Li is a researcher whose work lies at the intersection of computer vision, robotics, and autonomous systems, with a particular focus on 3D scene understanding, sensor calibration, and dynamic object modeling. His most cited paper, "SANet: Scene Agnostic Network for Camera Localization" (2019, 91 citations), introduced a groundbreaking neural architecture that decouples model parameters from specific scenes, enabling online, scene-agnostic camera localization—a significant departure from prior methods that required per-scene training. This work has been highly influential in advancing practical, real-time localization for augmented reality and robotics. Li also made notable contributions to multi-modal sensor fusion in "Single-Shot is Enough: Panoramic Infrastructure Based Calibration of Multiple Cameras and 3D LiDARs" (2021, 32 citations), proposing a single-shot calibration method that dramatically simplifies the integration of cameras and LiDARs, a critical step for mass production in autonomous vehicles. Earlier, his work on "Mobility Fitting using 4D RANSAC" (2016, 12 citations) tackled the challenge of capturing articulated motion from noisy, sparse dynamic data, offering a robust framework for modeling human and robotic movement. Across these contributions, Li has demonstrated a talent for solving practical, deployment-oriented problems in perception and sensing.