Mie University Students Challenge Real-World Data Analysis and Exercises Using Edge Devices in Data Scientist Development Program

What is the curriculum and goal for the Data Scientist Development Program?
Mie University offers practical educational programs to develop future leaders in data science. This is the fifth year of the program and is part of the collaboration between Nagoya University and two other universities. Mie University's Graduate School of Regional Innovation Studies takes the lead and conducts this program independently.

The point of the course at Mie University is to encourage students to think through a series of processes, by using real data from the world around them and using that data to uncover and solve problems. Students will experience difficulties in examining data in the real world, so we wanted students to learn practical skills, especially, like privacy issues, to know how important it will be in their careers to set their tasks and use data to solve them.

What were some of the tasks that were involved in the course?
Because we started the course by discussing task setting, it took a long time to determine our starting point, due to the high degree of flexibility in the program. In addition, for the first time in the program, we installed edge devices (AI cameras) provided by SSS at the university to leverage realistic data, which took extensive coordination with the university around the privacy considerations. AITRIOS from SSS and EBILAB explained that the image data does not leave the AI cameras because it is processed directly on the device. We were able to make additional adjustments within the cameras to complete the installation for data acquisition while preserving privacy.

What are some of your takeaways throughout the course term?
In previous courses, we utilized a variety of tools for data collection, model training, and inferencing. In this program term, SSS provided an integrated environment with their AITRIOS software and AI cameras, which enabled students to experience the entire machine learning development process within a single ecosystem, making the experience much more valuable. In the future, we will need to examine the problems and construct hypotheses about what we can do to solve them. I hope that students will understand that they can construct hypotheses based on ideas formed from the data.

Three Mie University students

Could you describe what you focused on in the course?
We exchanged opinions with related parties on what kind of challenges are most prominent on the university campus and adjacent Mie University Medical School Hospital and deepened our understanding of the issues through field visits. With that information, we narrowed down the themes and locations of edge devices (AI cameras) and factored in installation conditions.

In the end, we decided to concentrate on two themes to focus on for the program term:

The steps for the first use case were to set up a camera indoors so that the entire parking lot could be seen. We observed that cars parked in the parking lot were recognized as a group of cars rather than individual cars, so the congestion level per hour was expressed by the sum of the area of the group. As a flow, AITRIOS was able to acquire data, create AI models, and then analyze the data. Some parts of the image acquired by the camera were oblique, and the image was converted to an image from directly above by homography.

Could you tell us about what you learned from the program, and what you found challenging?
I learned that there are a lot of things to consider when acquiring data in the real world, including privacy concerns. Before I built AI models, I thought that public data sets would be accurate to a certain extent, but I realized that learning on real data was important because I didn't get any accuracy when I did it.

However, since this course was only about six months long, I felt that I didn't have enough time to learn all there is to know. To solve a big problem, I felt that I needed more time to accumulate and analyze data on a multi-year basis. But overall, I was very impressed and surprised that I was able to get this far with just one camera!

Would you recommend this course to someone you know?
Yes. It was a very good program in that even inexperienced people could participate. I thought it would be very difficult to create AI, but it was good that I could do it easily while using AITRIOS.

After taking this course, I think the barriers at the starting line when I enter the workforce will be lower and support me in finding a job related to AI and IoT. My interest in AI and IoT is stronger than before I took the course, and my future career is easier to visualize.