Jagan Charjee Pyaraka – A Minimal Key-Point Learning Framework for Robot Skill Transfer via HOI Recognition and Task Segmentation

This presentation will introduce my PhD research, “Minimal Key-Point Learning for Robot Skill Transfer from Videos.” I will begin by discussing key challenges in robotics, such as data scarcity, high computational costs, and the difficulty of adapting to new tasks. Next, I will outline my approach, explaining how minimal key-point tracking and neural networks enable robots to learn from human demonstrations. I will share my progress in HOI recognition and ongoing work on 6-DOF object tracking. Finally, I will discuss my planned research on task segmentation and robot execution, and how this work can enhance real-world robotic applications.

Jagan Charjee Pyaraka, a PhD researcher at Swinburne University of Technology, is part of the Biomimic Cobots Program at the Australian Cobotics Center, where he is focusing on developing lightweight and efficient methods for robotic skill transfer. His academic journey began in electrical and electronics engineering, later transitioning to robotics and evolving into a deep exploration of computer vision, machine learning, and human-robot interaction. He has contributed to a variety of academic and industry projects, spanning QA automation, intelligent system development, and AI-driven robotics. Before pursuing his PhD, he gained industry experience as a Senior QA Automation Engineer at NTT DATA Services, refining his expertise in system optimization and automation frameworks.

Motivated by AI’s potential to revolutionize human-robot collaboration, his research focuses on developing a minimal key-point learning framework for efficient robotic skill transfer. His work integrates HOI recognition, meta-learning, and task segmentation, advancing the field of resource-efficient robotic intelligence.