OKI Technical Review

2019 No.233 AI for Human-centric Society

Model Pruning Technology for Deep Neural Networks [191KB] PDF

Kohei Yamamoto, Motoko Tachibana, Kurato Maeno

The deep learning model, which is a core AI technology, has been difficult to implement in small devices due to its high operational complexity. This article introduces current status and problems associated with the compression technologies used to reduce operation while maintaining the original accuracy, and presents OKI's own unique technology.

Failure Prediction for Multi-function Machine based on Vibration Data Analysis [421KB] PDF

Kazuki Kobayashi, Masatoshi Sekine

OKI has developed an anomaly detection method applicable to a multi-function machine that is difficult to handle with conventional vibration analysis method. This article provides an overview of the developed method and the results of an evaluation experiment conducted on an actual machine, which OKI produces.

Abnormal Behavior Detection using Image Sensing [749KB] PDF

Reiko Kishi, Phan Trong Huy, Kazuma Yamamoto, Makoto Masuda

Possible applications of abnormal behavior detection using image sensing are presented, and as an actual example, OKI's effort in developing a technology that combines abnormal behavior detection with information obtained from an ATM is introduced.

Reports on misuse of public research funds and misconduct related to research
activities can also be filed from the page that opens when you click the above button.

Special Contents