Hikaru's avatar

Hikaru TANABE

Graduate Student at UEC Tokyo
Design Thinking and Data Science (Dx2) Program

EmailGitHubLinkedInTwitter

About me

田邊光（たなべひかる）と申します。電気通信大学大学院（UEC）の博士前期課程に所属し、Vision & Language 分野の研究や Web サービス開発に取り組んでいます。最近は vision-centric なタスクを解ける MLLMs に特に関心を持っています。


Specializing in Vision-and-Language research and web development. My work focuses on enhancing food understanding through Multimodal Large Language Models (MLLMs), with an emphasis on reasoning, segmentation, and 3D-aware nutrient estimation. I am particularly interested in vision-centric tasks using MLLMs, and broadly in exploring the potential of multimodal intelligence.

Interested / Working Topics

Multimodal / Agentic LLMs, Vision & Language, Food Image Recognition, Embodied AI, Web Development

Publications

  • Hikaru Tanabe and Keiji Yanai, Decoupled Clip and Dynamic Sampling Policy Optimization for Food Reasoning Segmentation, MMFood@ACMMM2025, Dublin, Ireland, (2025/10).
  • Mayu Ogishi, Hikaru Tanabe and Keiji Yanai, Diffusion-Guided 3D-Aware Calorie Estimation from a Single Food Image, MMFood@ACMMM2025, Dublin, Ireland, (2025/10).
  • Yuto Nakamizo, Ryuhei Miyazato, Hikaru Tanabe, Ryuta Yamakura and Kiori Hatanaka, Ensembling Multiple Hallucination Detectors Trained on VLLM Internal Representations, KDDCup 2025 Workshop for META CRAG, Toronto, Canada, (2025/08).
  • Hikaru Tanabe and Keiji Yanai, Reasoning-Driven Food Energy Estimation via Multimodal Large Language Models, Nutrients (2025/03).
  • Hikaru Tanabe and Keiji Yanai, CalorieVoL: Integrating Volumetric Context into Multimodal Large Language Models for Image-based Calorie Estimation, MMM2025, Nara, Japan, (2025/01).
  • Hikaru Tanabe and Keiji Yanai, CalorieLLaVA: Image-based Calorie Estimation with Multimodal Large Language Models, MADiMa@ICPR2024, Kolkata, India, (2024/12).
  • 松浦史明, 中溝雄斗, 田邊光, 柳井啓司, BlipNeRF: MLLMによる不可視視点の予測を用いた少数の物体正面画像からの3次元再構成, 画像の認識・理解シンポジウム (MIRU2025), Kumamoto, Japan, (2025/07).
  • 大岸茉由, 田邊光, 柳井啓司, 拡散モデルベースの３次元復元手法を用いた単一食事画像からのカロリー量推定, 画像の認識・理解シンポジウム (MIRU2025), Kumamoto, Japan, (2025/07).
  • 田邊光, 柳井啓司, 大規模視覚言語モデルと食品体積推定による食事画像からのカロリー量推定, 画像の認識・理解シンポジウム (MIRU2024), Kumamoto, Japan, (2024/08).
  • 田邊光, 柳井啓司, 大規模視覚言語モデルを用いた食事画像からのカロリー量推定, CVIM 2024年5月研究会, Tokyo, Japan, (2024/05).

Experience

  • Algomatic Co., Ltd. - ML Engineer Internship

    Tokyo, Japan | 2023/10 - Present

  • Teaching Assistant (Educational Computer System Management)

    The University of Electro-Communications, Tokyo, Japan | 2025/04 - 2025/08

  • Teaching Assistant (Lecture: Media Science and Engineering Laboratory)

    The University of Electro-Communications, Tokyo, Japan | 2024/12 - 2025/02

  • NEC Laboratories America, Inc. - Research Internship

    San Jose, CA, USA | 2024/08 - 2024/10

  • Teaching Assistant (Lecture: Object Recognition)

    The University of Electro-Communications, Tokyo, Japan | 2023/10 - 2024/02

  • Lograph Inc. - Web Developer Internship

    Tokyo, Japan | 2022/08 - 2023/12

  • Yamazaki Lab. - Research Assistant

    The University of Electro-Communications, Tokyo, Japan | 2022/02 - 2022/08

Education

  • MEng in Design Thinking and Data Science

    The University of Electro-Communications, Tokyo, Japan | 2024/04 - Present

  • BEng in Media Science and Engineering

    The University of Electro-Communications, Tokyo, Japan | 2020/04 - 2024/03