Max Robotics — Global Service Robot Directory + US Certification

About

Qingsong Yan is a computer vision researcher whose work sits at the intersection of stereo vision, depth estimation, and indoor robotics perception. His most influential contributions center on the development of large-scale stereo datasets designed to train deep learning models for scene understanding tasks, particularly disparity and surface normal estimation. His flagship project, the Indoor Robotics Stereo (IRS) dataset, has become a notable resource in the field, appearing in multiple iterations between 2019 and 2021 and accumulating over 57 citations combined — reflecting its sustained value to researchers building systems for robotic localization, navigation, and interaction. By bridging the gap between synthetic and naturalistic training data, Yan has helped advance the robustness of deep models in real-world indoor environments. More recently, he has expanded his research scope to tackle the challenging domain of panorama depth estimation, addressing longstanding issues of geometric distortion and projection discontinuity through novel approaches such as SphereDepth and SphereFusion. This trajectory demonstrates a consistent commitment to enabling richer, more geometrically accurate environmental perception — capabilities that are foundational to the next generation of autonomous robots and immersive visual systems.