Researching cutting-edge image recognition with high versatility. Creating new technologies through trial and error.

Profile

Graduate of Keio Shonan Fujisawa Senior High School, Kanagawa Prefecture

After a high school career dedicated to tennis, he pursued his strength in the sciences. At the Keio University Faculty of Science and Technology, he took advantage of the *gakumon* system to study a wide range of subjects while searching for his true passion. An experiment in programming techniques using images sparked his enjoyment of making things and achieving desired results, which heightened his motivation for research. He immersed himself in the study of image recognition, a field with promising future applications, and participated in academic conferences both in Japan and abroad. He is dedicated to his research, expanding his ideas through interactions with various researchers and pursuing novel technologies.

[*] Academic year at the time of the interview (July 2019).

Since I was admitted through internal advancement and didn't have to study for entrance exams, I devoted myself to club activities. I was on the tennis team, and we practiced a lot, including a 5 km dash every morning. Even after classes, I trained until late. My high school years were the busiest and toughest, but my fondest memory is leading the team as vice-captain to our school's first-ever appearance in the Inter-High School Championships in our final team competition. Also, although my club activities were demanding and I couldn't study as much as I wanted, I found time to teach myself subjects I enjoyed, such as physics and computer science.

I've loved numbers since I was a child (laughs). My parents told me a story about how I used to read out numbers posted on the wall when I was little. While I was good at math, I was quite poor at Japanese language... Although I studied hard for Japanese during my junior high entrance exam preparations, I never grew to like it, so I chose the Faculty of Science and Technology without hesitation. I had no other options besides science, but since I wasn't sure what I wanted to do, entering the Faculty of Science and Technology with its *gakumon* system [*1] was a reassuring choice.

[*1] The *gakumon* system: A system where students choose one of five *gakumon* (academic clusters) at the time of admission. After enrolling, they gradually narrow down their field of study according to their interests and concerns, and decide on their department when they advance to their second year.

Before enrolling, I thought the Faculty of Science and Technology would be much busier. As a high school student, I used to leave home at 5:30 every morning, so my mornings felt more relaxed. Of course, my first and second years were busy with many required courses, but it felt like high school classes, and I think I had a good time with everyone. By my second and third years, I got used to the freedom and realized that I wouldn't gain anything just by passively attending classes. To find what I wanted to do, I made an effort to use my time effectively, such as by studying programming on my own. My impression is that many of the friends I made in the Faculty of Science and Technology are involved in various things, not just studying, but also excelling in sports or their hobbies.

Until I joined my research lab, my days were all about tennis! I joined the "Faculty of Science and Technology Athletic Association Tennis Club (Rikotai Tennis Club) [*2]" and also served as captain. Since it's a tennis club tailored to the Faculty of Science and Technology's curriculum, it aims for a balance between academics and sports, and we practice on a relatively flexible schedule. That said, I think there were times when I couldn't manage both and my studies suffered (laughs). What's most memorable is achieving the best result in my final year. There's a league tournament where universities from the Kanto region gather. We were originally in the first division, but we were relegated to the second division after losing in the Waseda-Keio rivalry match in my first year. I felt frustrated for a long time. Normally, you retire from club activities in your third year, but I played in a match in my fourth year and we were able to return to the first division. Moreover, it was the Waseda-Keio rivalry match, so my efforts paid off, and I was thrilled. Even now, I sometimes visit the tennis courts to play or to coach. The appeal of tennis is that it simply feels good to hit the ball, and you can stay healthy while having fun.

[*2] Faculty of Science and Technology Athletic Association: The Faculty of Science and Technology Athletic Association, commonly known as "Rikotai," is an organization with characteristics between an official athletic team and a student club. It aims to balance academics and sports. Although it bears the name "Faculty of Science and Technology," Keio students from other faculties can also join.

I was originally torn between the Department of Physics and chose "Gakumon 1 [*3]." At the time, I was interested in statistical mechanics within physics, and as I studied it, I found many mathematical formulas that connected to the information field. Statistical mechanics can be applied to semiconductors, and I was confident in my programming skills, so I felt that the Department of Electronic Engineering (now the Department of Electronics and Electrical Engineering) was the place to combine and utilize these various skills. I'm currently researching a field called computer vision, and the results are visually clear, like using a camera to recognize and output an image. It's also appealing because it's simply more fun to look at than just numbers and graphs. The Department of Electronic Engineering (now the Department of Electronics and Electrical Engineering) is a department where you can study a wide range of subjects, from optics to semiconductors and information science. Compared to other departments, it offers a high degree of schedule flexibility, allowing you to focus on the fields you want to pursue and engage in the studies you enjoy.

[*3] Gakumon 1: The "Gakumon" that, at the time of his 2015 admission, allowed students to advance to one of four departments: the Department of Physics, the Department of Applied Physics and Physico-Informatics, the Department of Electronic Engineering (now the Department of Electronics and Electrical Engineering), or the Department of Mechanical Engineering. The names and composition of each *gakumon* were changed for students admitted in fiscal year 2020, and the "Department of Electronic Engineering" was renamed the "Department of Electronics and Electrical Engineering" in fiscal year 2020.

The Robot Vision Programming course I took in the fall of my third year, when my studies were becoming more specialized. It was an experiment taught by Professor Yoshimitsu Aoki, whose lab I'm currently in. It spanned six weeks, where we attached a webcam to a computer and learned various image processing techniques. At first, we were given an image and taught how to do things like move a soccer ball freely within it, or extract a person from a video and replace the background with a white one. Then, we used the techniques we learned to actually develop an application. Our group created an app that recognizes human movement with a camera and produces sound. The mechanism was that moving your hands in the air would play an instrument according to the movement, and for our presentation, we used it to play "Kaeru no Uta" (The Frog Song). Through this highly creative experiment, similar to what I'm doing in the lab now, I was able to experience the fun of making things and the joy of achieving results.

I belong to the Aoki Lab, where we research image recognition, also known as computer vision. Within that, I am conducting research on Vision & Language, which combines images and language. It's research where you input an English instruction for an image, and an image is generated that has been transformed according to the instruction. For example, if you have an image of a man with black hair and simply input "make his hair blond," it will output the image with him having blond hair. Other projects include creating a program that predicts unseen areas from a photo taken from a single direction to generate a 360-degree image, and research that extracts information from video to analyze what actions the people in it are actually performing. It's a field where practical application is still on the horizon, but as computer vision develops, we can expect a wide range of applications, such as improved accuracy in image searches, applications in AI, and in the sports field, automatically judging the number of rotations in a figure skating jump or analyzing plays in rugby from video.

The lab is very active in discussions, and we are vigorously engaged in creating and disseminating valuable work. The pace of attending conference presentations and publishing papers is fast, creating a stimulating environment where I can focus on my research.

Participating in a program for young researchers at a conference called "MIRU 2018: The 21st Meeting on Image Recognition and Understanding Symposium [*5]." There was a project aimed at acquiring new skills, where we had to investigate and present on papers from a field different from our specialty of image recognition. I collaborated with researchers of my generation on this investigation. We started having meetings on Skype about two months in advance, read and summarized the papers, and presented at the main conference. As a result, our group was able to win an award. Also, the paper on Vision & Language that I read at that time became the catalyst for my current research. Since then, I've had more opportunities to meet researchers from other universities at other conferences and study groups, and my community has expanded.

I recently attended a top-tier international conference in Los Angeles, and it was amazing to be able to freely discuss with the world's top researchers and gain a lot of information and ideas. Next year, I want to engage more with researchers from overseas through my papers.

[*5] MIRU 2018: The 21st Meeting on Image Recognition and Understanding Symposium: Japan's largest academic meeting on image recognition and understanding technology, where researchers, engineers, and students present and discuss the latest research, from basic to applied, every year.

After graduation, I'm thinking of working for a private company. Image recognition is a hot field, so it would be great if I could make a career out of it, but I'm torn between a research position and a development position. I really enjoy research, but it's also a field where research results are easy to implement, so I'm also interested in software and product development. I want to try creating something interesting that has an impact on many people and that they find useful. Currently, through a connection with a lab alumnus, I'm able to conduct research freely as a part-time employee at a company twice a week, so I'm hoping to participate in an internship this summer where I can be involved in development work.