Master thesis in Physics presentation. The title is "Learning-Based Human Anthropometric Parameter Estimation from Images".
This thesis presents a multi-modal framework for estimating human height and weight from a single RGB image. The proposed method combines human segmentation, monocular depth estimation, and vision–language model predictions through a learnable fusion module to improve robustness and accuracy. The framework is further validated in both a digital twin simulation environment and a robot-to-human object handover task, demonstrating its potential for personalized robotics and human–robot interaction applications.
Supervisor: Karinne Ramirez-Amaro (Electrical Engineering, Chalmers)
Opponent: Yihuai Cai
Zoom: