This is my bachelor thesis project.
In Japanese agriculture, green soybeans are graded according to the Japanese cooperative standards before being dispatched to retailers such as supermarkets and grocery stores. This sorting task is still done manually, which results in significant overwork for farmers in Japan. In this study, we focused on the automation of the grading tasks using AI. This research involved a mixture of computer vision and the use of a single-short classification deep neural network, proposing a method for collecting images of green soybeans from multiple angles including top and side angles. Using the custom dataset, we enhanced AI model performance, achieving a grading accuracy of 93%, which exceeds previous studies.