Graduate School of Keio University

School of Integrated Design Engineering
Center for Electronics and Electrical Engineering

April 2019 – Present Japan
My research theme is about human motion analysis and object function detection. I’m expected to get a master’s degree in 2021.
Adviser: Prof. Yoshimitsu AOKI
Lab. Homepage

Keio University

Dept. of Electronics and Electrical Engineering

April 2015 – March 2019 Japan
I mainly studied Electronics in the first three years. During the last year, I researched Machine Learning, Computer Vision and Robot Vision.
Adviser: Prof. Yoshimitsu AOKI
Lab. Homepage




SoftBank Corp.

Aug 2019 – Sep 2019 Tokyo, Japan
Worked on two task:
1) classifying a product into normal one or abnormal one and visualizing where a CNN model looks.
2)semantic segmentation for super high-resolution images

Research Assistant

National Institute of Advanced Industrial Science and Technology(AIST)

Apr 2019 – Present Tsukuba, Ibaraki, Japan
Research about Machine Learning and Action Recognition under the supervision of Ph.D. Hirokatsu KATAOKA.


IBM Japan, Ltd.

Sep 2018 – Mar 2019 Tokyo, Japan
Worked on weakly-supervised affordance detection using the hierarchy between affordances and objects.



  • 石川裕地,石川晴也,秋月秀一,青木義満,操作タスク入力に基づく物体の機能部推定, 精密工学会 85巻12号 (2019年12月5日発行)

International Conference

  • (Under Review)
  • Yuchi Ishikawa, Haruya Ishikawa, Shuichi Akizuki, Masaki Yamazaki, Yasuhiro Taniguchi, Yoshimitsu Aoki “Task-oriented Function Detection Based on Operational Tasks” in Conference: 2019 19th International Conference on Advanced Robotics (ICAR). (Acceptance Rate 55.3%)
  • Seito Kasai*, Yuchi Ishikawa*, Tenga Wakamiya, Kensho Hara, Hirokatsu Kataoka, “AIST Team submission for Task 3, Dense-Captioning Events in Videos,” CVPR 2019 Workshop, International Challenge on ActivityNet Challenge, 2019.
  • Tenga Wakamiya, Kensho Hara, Yuchi Ishikawa, Seito Kasai, Hirokatsu Kataoka, “AIST Submission for ActivityNet Challenge 2019 in Trimmed Activity Recognition (Kinetics),” CVPR 2019 Workshop, International Challenge on ActivityNet Challenge, 2019.
  • Haruya Ishikawa, Yuchi Ishikawa, Shuichi Akizuki, Yoshimitsu Aoki, Human-Object Maps for Daily Activity Recognition, International Conference on Machine Vision Applications(MVA 2019)

Domestic Conference

  • Seito KASAI, Yuchi ISHIKAWA, Tenga WAKAMIYA, Kensho HARA, Hirokatsu KATAOKA, Exploring the Best Model for Dense Captioning Events in Videos, 第22回画像の認識・理解シンポジウム (MIRU2019)
  • 若宮天雅,原健翔,石川裕地,笠井誠斗,中村 明生,片岡 裕雄,超大規模データセットによる動画像認識のための学習済みモデルの構築,第22回画像の認識・理解シンポジウム (MIRU2019)
  • 石川裕地,石川晴也,秋月秀一,青木義満,ロボットの物体操作のためのタスク指向な機能部の推定,第22回画像の認識・理解シンポジウム (MIRU2019)
  • 秋月秀一,石川裕地,石川晴也,青木義満,物体の配置記述に基づくシーン復元のための操作方法推定,第25回画像センシングシンポジウム(SSII 2019)
  • 石川裕地,石川晴也,秋月秀一,青木義満,操作タスクと指示対象クラスの入力による物体の機能部推定,第25回画像センシングシンポジウム(SSII 2019)
  • 石川裕地,石川晴也,秋月秀一,青木義満,操作タスク入力に基づく物体の機能部推定,動的画像処理実利用化ワークショップ2019(DIA 2019)
  • 秋月秀一,石川裕地,石川晴也,青木義満,非定常状態の理解とシーン復元のための物体の操作方法推定,動的画像処理実利用化ワークショップ2019(DIA 2019)


CVPR2019 速報

I took part in CVPR 2019 and wrote this article with members of cvpaper.challenge.


As a member of cvpaper.challenge, I read a lot of papers accepted at CVPR or several top conferences. I also reseach and share the knowledge with its members.

ActivityNet Challenge

Out team took part in ActivityNet Challenge in CVPR workshop. We won 9th place in Task A - Trimmed Action Recognition. Our team also participated in Task 3 - Dense-Captioning Events in Videos, though the result was not good.

Commissioned Research by a chemical manufacturer

Work on Semantic Segmentation for specific objects as a contract research.

Paper Summary

I read papers every day and summarize them as far as possible in GitHub. If you get interested, visit my project page.

Collaborative Research with a leading automaker

Research how to make robots to understand the way of using objects with a major automaker in Japan.


JDLA Deep Learning for GENERAL

JDLA aims to develop Deep Learning Generalist, capable of utilizing in business, which has sufficient knowledge in Deep Leaning.

Outstanding Performance Award for Bachelor Theses

My bachelor thesis topic is “Task-oriented Function Detection Based on Operational Tasks”.
I proposed an alternative representation to Affordance, “Task-oriented Function”, in the paper. This representation makes it possible to desribe a variety of ways to use an object, though only one usage can be described in Affordance Detection in the field of Computer Vision.

3rd place award in MIRU 2018 Young Researchers Program

As a young researcher program in MIRU 2018, participants were divided into several groups and each group read papers in the field outside Computer Vision. Then, each group summarized the history, the trend and the conection with Computer Vision. It helped us not only understand different fields, but also consider how we can make use of knowledge about them for Computer Vision.

Our group, C, read papers in the field of Robotics, focusing on “Transferring Knowledge from Simulation to Real”. You can see our poster and presentation material from this link.





I received this scholorship which aims at supporting up to a hundred of capable students studying Artificial Intelligence in Japan.



Python / C / C++ / MATLAB / HTML / CSS

Computer Vision

Human Motion Analysis / Action Recognition
Semantic Segmentation / Object Detection


Statistics / Linear Algebra


Japanease / English





When I’m Not Programming

When I’m not in the lab., I put my free time into my hobbies or hang out with my friends.
When I take a long vaction, I always do travel all over Japan or the world. I love traveling and I’ve experienced backpacking many times such as traveling across Japan by bicycle, crossing America by motor vehicle, hitchhiking, survival on an uninhabited island, camping in Africa and so on. I’ve been to about 25 countries.

My hobbies are: