深層学習を用いたブラウザ環境下で非接触操作を提供する汎用ソフトウェアモジュールの提案

抄録

In this paper, the authors report on the development of creating a software module that can make arbitrary web page have contactless operation feature with neural network model. To obtain hand coordinate data for training, the authors created a special web application with MediaPipe which is an open-source library provided by Google. Thanks to the application, sufficient number of data can be obtained easily. As a result, the authors confirmed that the model can provide excellent UX. We then experimented to evaluate the performance of a software module used the model and confirmed that how accurately it could work with that module.