"What I do" in 30s, 5mins, and 10mins
- "What I do" in 30s, 5mins, and 10mins
- 30 seconds
- 5 minutes
- 10 minutes
- What I would like to accomplish
Hi, I'm currently researching individual & group human motion at the University of Tsukuba🤸
My interests intersect sports science, computer vision and agent based simulations.
Using cutting edge tech to help sports teams and athletes up their game is not only fun, the by-products can advance our understanding of how our bodies work, strategies, game theory, tracking algorithms and so on!
Currently, I am working on three projects.
- Multi-agent Learning for Football The main goal of this research project is to build a framework that lets us learn football from self learnt agents. Alpha Go showed us humans that we aren't the best at playing board games! Although we might still be able to outperform RL in football, I think there still can be a lot to learnt from how they play.
- Analyzing Football Games with Computer Vision The main goal of this research project is to build the basic building blocks necessary for analysis using computer vision. I think tracking is one of the most difficult aspects in sports analytics and it is preventing many amatuer teams from benefitting from numbers. My colleague Ikuma and I are working on many problems in this area. ⭐ Our full paper (Co-authored, equally contributed) has been accepted by ACM MM Sports 2021! "Automated Offside Detection by Spatio-Temporal Analysis of Football Videos"
- Analyzing Sports Actions with Computer Vision The main goal of this project is to quantitatively analyze actions in sports using videos. This is important because it gives everyone a way to measure and improve their actions. ⭐ We are currently building an app to measure the speed of a kick football: KICK Pro. The beta release will be available this october!
I think I have time to introduce a little background here.
I have always been passionate about data analysis in football. I spent my undergrad days as an data analyst (while competing in the b-teams as an athlete) at the University of Tsukuba. We became first in the national tournament and several teammates went on to play in the J League (as a consequence got some media attention too 😉).
Although having experienced some success back then, I must admit most of it was due to lack and the sheer excellence of the players and coaches. To be honest, any sophisticated analysis is still in its infancy and has a long way to go until it can be usable in the mainstream.
That is why I decided to decline any dream job offers and am currently pursuing a masters degree. At first I was doubtful of my decision but with some people gradually starting to take interest in the intersection of sport and A.I, I don't think I made the wrong choice.
Advancing sports analytics through AI research
Advancing sports analytics through AI research Creating testing environments to help progress AI research out of the lab and into the real world is immensely challenging. Given AI's long association with games, it is perhaps no surprise that sports presents an exciting opportunity, offering researchers a testbed in which an AI-enabled system can assist humans in making complex, real-time decisions in a multiagent environment with dozens of dynamic, interacting individuals.
Google Research Football with Manchester City F.C.
Train agents to master the world's most popular sport
What I would like to accomplish
I think we have a long long long long way to go to gain significant insight on gaining meaningful information from vision or reinforcement learning, but I would like to see it happen and make bank on it in my life time!
Know how to answer "what do you do" in 30s, 5mins, and 10mins
Often in my department, we give demos of our work to outsiders: students, potential students, parents, employees, visitors, you name it, we've probably given them a demo. In one of these situations, I started to realize that me and my colleague (another grad student) didn't have enough stuff to fill the time for our demo...