Max Robotics — Global Service Robot Directory + US Certification

About

Beiwen Tian is a researcher at the forefront of robotic vision and 3D scene understanding, with a focus on making perception systems more data-efficient and robust. Their work addresses critical bottlenecks in point cloud analysis, particularly the challenge of annotation scarcity. Tian's major contributions include pioneering semi-supervised and omni-supervised room layout estimation from point clouds, a task essential for robot environment sensing and motion planning, achieving 14 citations for their foundational 2023 paper. They further advanced the field with pointly-supervised 3D scene parsing using a viewpoint bottleneck, demonstrating how to learn from extremely sparse labels. In the domain of embodied AI, Tian introduced TOIST, a task-oriented instance segmentation transformer that goes beyond noun-based detection to understand verb-referenced actions—a crucial step for robots interpreting human intent. Most recently, their work on ImitDiff leverages foundation-model priors to create distraction-robust visuomotor policies for robotic manipulation. By tackling data scarcity and semantic understanding across multiple tasks, Tian's research is shaping more capable and practical robotic vision systems for real-world deployment.