Max Robotics — Global Service Robot Directory + US Certification

About

Gurkirt Singh is a computer vision researcher whose work sits at the intersection of action detection, robotic perception, and video understanding. His research primarily focuses on developing systems capable of identifying and localizing human actions in video — a technically demanding challenge with far-reaching applications in robotics, autonomous driving, and surgical assistance. Singh's most notable contributions include pioneering work on incremental tube construction for real-time human action detection, addressing a critical gap in systems that previously relied on offline batch processing. His Two-Stream AMTnet framework advanced video-based action representation by integrating temporal modeling with online tube generation. Particularly impactful is his development of the ESAD (Endoscopic Surgeon Action Detection) dataset, which has garnered 29 citations and represents a landmark resource for training surgical assistant robots to recognize and anticipate surgeon movements during live procedures — a contribution with meaningful implications for patient safety and autonomous surgical systems. Singh has also extended action detection to autonomous vehicle contexts through the READ dataset, demonstrating the versatility of his methodological expertise. With work spanning surgical robotics, human-robot interaction, and self-driving perception, his research consistently bridges foundational computer vision techniques with high-stakes real-world applications, establishing him as a notable contributor to applied action recognition research.