Enhancing Robot Perception with Real-World HRI

Robot perception often fails in uncontrolled environments due to unfamiliar object classes, different domains, or hardware issues. This poses significant challenges for human-robot interaction outside of a lab or user study settings. My work focuses on two separate approaches: improving robot perception models and developing systems where users can directly correct robot errors. My research strives to improve HRI in real-world scenarios by reducing vision errors and empowering users to address them.