Proposal of eye-camera video analysis method using Mask R-CNN

Abstract

In recent years, in Japanese agriculture, the opportunities for new farmers to receive direct guidance from skilled farmers have decreased due to the declining working population and the aging population. Along with this, the maintenance of agricultural work manuals has become more important. Therefore, this research group has proposed a web-based farming technology inheritance manual using an eye camera.However, the proposed manual had a problem that the burden of preparation was large. The purpose of this study is to solve that problem. In this study, we implemented a system that automatically transcribes what a skilled farmer was looking at then using an eye camera and Mask R-CNN, and evaluated its performance. As a result, we believe that the burden of creating the proposed manual can be solved.