User interview: Nihon University

Nihon University has used ADFI for research and development of inspection systems. We interviewed Masato Iguchi, an assistant professor at Nihon University’s College of Science and Technology (March 10, 2023), and asked him about the use of ADFI in his laboratory.

Q: Please give us an overview of your laboratory.

A: Approximately 20 students belong to our laboratory, and they are researching several themes related to building facilities such as optimization of maintenance plans for building facilities. For example, we are researching how to automate the inspection and maintenance of air conditioning equipment such as air conditioners.

Q: How are you using ADFI in your laboratory?

A: This fiscal year, we are using ADFI for two research themes related to automating inspections of air conditioning equipment such as outdoor units of air conditioners. Specifically, we are using ADFI to create an AI that detects equipment abnormalities.

Q: Please give us an overview of the first research theme.

A: This is research on a system that automates the inspection of outdoor unit surfaces. This system uses a camera to check whether foreign matter is attached to the outdoor unit. In the experiment, foreign matters under various conditions were attached to the surface of the outdoor unit and were captured with a camera from various angles. We evaluated whether the AI could detect foreign matters from the images.

Q: Please give us an overview of the second research theme.

A: This is research to detect failures and abnormalities in outdoor units from operating sounds. We installed an industrial stethoscope in the outdoor unit, converted the acquired sound into a spectrogram image, and conducted an experiment to detect anomalies from the image with an AI.

Q: What is your impression of using ADFI?

A: ADFI is great, because you can create AI immediately with intuitive operations on a web browser. Since our laboratory is in the Department of Architecture, there are many students who are not familiar with information engineering, but they quickly got used to the operation of ADFI and were able to use AI in their research. Also, the output results of AI created with ADFI are more accurate than I expected, which surprised me.

Q: What kind of people would you recommend ADFI to?

A: I would recommend it to anyone who wants to automate foreign matter detection or anomaly detection. Intuitively, it seems that ADFI can distinguish quite a variety of objects, so in many cases, foreign matters can be detected with high accuracy. Also, I would like people like the students in our laboratory who are not involved in AI development to give ADFI a try.

* All information (organization names, affiliations, titles, content of interviews, etc.) is as of March 2023 (at the time of the interview).