Shuhei Kurita

About

Shuhei Kurita is at the forefront of embodied AI and human-robot interaction, with a research focus on grounding natural language in first-person visual perception. His most impactful contribution is the creation of **RefEgo**, a pioneering dataset for referring expression comprehension from egocentric views, which has garnered 11 citations since 2023. This work addresses a critical challenge in developing context-aware agents—such as smart glasses and autonomous robots—that can interpret intuitive text instructions to localize objects in their surroundings. Kurita further advances embodied intelligence through his work on **zero-shot Embodied Question Answering (EQA)**, where he introduced a map-based modular approach that enables robots to navigate novel environments and answer human queries without prior training. He also explores the validation of **LLM-generated object co-occurrence information** to enhance robots’ spatial understanding of real-world 3D scenes. By bridging large language models with robotic perception, Kurita’s research pushes the boundaries of how machines comprehend and act upon human language in dynamic, first-person contexts—paving the way for more intuitive and capable assistive technologies.

Research Focus

Key Achievements

Top Papers

1. 1
   RefEgo: Referring Expression Comprehension Dataset from First-Person Perception of Ego4D

   11 citations · 2023
2. 2
   Map-based Modular Approach for Zero-shot Embodied Question Answering

   3 citations · 2024
3. 3
   Validation of LLM-Generated Object Co-Occurrence Information for Understanding Three-Dimensional Scenes

   2 citations · 2024
4. 4
   RefEgo: Referring Expression Comprehension Dataset from First-Person Perception of Ego4D

   2 citations · 2023

Key Collaborators

Contact & Links