Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 05, 2019 - June 08, 2019
In Japan, the number of casualties due to an industrial accident in 2009 was 114,154 people. Currently, a comprehensive system to prevent such an accident is basically done manually. Furthermore, it has not been automated. Therefore, in this study, we develop an anomaly detection method for the prevention of industrial accident using machine learning technology. In order to carry out anomaly detection reliably, we use a skeleton map of a person as training data from skeleton information extracted by OpenPose. Here, Variation Autoencoder (VAE) is applied as a deep learning model. We confirm that detection results with high accuracy are produced compared with the conventional method.
