Upskilling Education of AI Literacy and Building Implementable AI for Edge Computing System Using Single Board Computer

Abstract

In recent years, artificial intelligence (AI) has attracted considerable attention not only in research fields but also in society. The performance of AI has dramatically improved on a daily basis in various areas such as language translation, image recognition, and information search. In industries, research and development are also underway to improve productivity and reliability. Visual inspection is one of the attracted topics for checking product anomalies in manufacturing lines and infrastructure defects. Therefore, the demand for engineers capable of developing AI-based solutions is increasing. In this situation, Tokushima University began an upskilling course of AI (AI-Kouza) for employees of local companies in 2018. The computer used in the lecture was a small single-board computer that was considered applicable for edge computing and allowed everyone to experience AI widely. The goal of the lecture was to acquire AI literacy and build an original AI model through exercises. To date, there is no literature on AI lectures with hands-on exercises starting from creating dataset that aims to implement AI in robots or edge-computing systems. This study pro-poses an upskilling education program, including the lecture concept and design, focusing on classification problems such as image recognition. Survey results collected from the participants in each lecture were analyzed to evaluate the program. From the results, they found it difficult to understand the mathematical theories of AI models and program codes. However, they were eager to understand the theory and concept of AI models as well as the higher technical skills required to build an original AI.