Max Robotics — Global Service Robot Directory + US Certification

About

Andrew Zisserman is a leading figure in computer vision, whose work bridges geometric reasoning, image segmentation, and multimodal perception. His research spans interactive image segmentation, visual navigation, and self-supervised depth estimation, with a recent foray into physics-informed audio-visual learning. Zisserman’s most cited paper, "Geodesic Star Convexity for Interactive Image Segmentation" (2010, 412 citations), introduced a powerful shape constraint that extended star-convexity priors using geodesic paths, enabling more accurate and user-friendly segmentation with global energy minima. Earlier, he explored visual navigation around curved obstacles (2002), using silhouette geometry to plan minimal paths for robots. More recently, his work on self-supervised instance adaptation for monocular depth estimation (2020, 20 citations) advanced robotic perception by learning depth from raw video without 3D ground truth. A notable achievement is his pioneering study on inferring physical properties from pouring liquids using sound alone (2025), demonstrating cross-modal reasoning. Zisserman’s contributions have profoundly influenced interactive tools, autonomous navigation, and self-supervised learning, making him a key innovator in vision and robotics.