Baorui Ma

Beijing Academy of Artificial Intelligence

Papers

Total Citations

H-Index

About

Baorui Ma is a rising researcher in computer vision and robotics, whose work focuses on the critical challenge of cross-modality registration—aligning 2D images from cameras with 3D point clouds from LiDAR sensors. This capability is fundamental for autonomous driving, robotics, and 3D scene understanding. Ma’s major contribution is pioneering differentiable and relational learning frameworks that enable robust, end-to-end registration between these disparate data types. Their 2023 paper, "Differentiable Registration of Images and LiDAR Point Clouds with VoxelPoint-to-Pixel Matching," introduced a novel method that bypasses traditional, non-differentiable Perspective-n-Point (PnP) solvers, achieving 9 citations and establishing a new paradigm for learning-based alignment. Building on this, their 2025 work, "RelaI2P: Relational Learning for Image-to-Point Cloud Registration," advances the field by modeling complex relationships between pixel and point features, moving beyond simple pattern matching. Ma’s research directly addresses a core bottleneck in multimodal perception, offering more accurate and efficient solutions for real-world systems. Their work is already influencing how autonomous vehicles and robots fuse visual and spatial data, marking them as a key innovator in this rapidly evolving domain.