@techreport{oai:ipsj.ixsq.nii.ac.jp:00220079,
 author = {久保, 智哉 and 有本, 和民 and 横川, 智教 and 穂苅, 真樹 and 茅野, 功 and 梶原, 景久 and Tomoya, Kubo and Kazutami, Arimoto and Yokogawa, Tomoyuki and Masaki, Hokari and Isao, Kayano and Kagehisa, Kajiwara},
 issue = {5},
 month = {Sep},
 note = {ドライバーモニタシステムにおいて，顔表情評定を AI システム上で動作させるためにニューラルネットワーク（NN）に 3 次元畳み込みニューラルネットワーク (3D CNN) を用いた研究があるが，車に搭載可能なマイコンボード上で動作させるには計算量が大きすぎる．本研究では，3D CNN から，2D CNN へコンパクト化することで，精度を保ちつつ，NN における計算量の削減を目指した．, In a driver monitoring system, a 3D convolutional neural network (3D CNN) has been used as a neural network (NN) to evaluate facial expressions on an AI system, but it is too computationally expensive to run on a microcontroller board that can be installed in a car. In this study, we aimed to reduce the computational complexity in the NN while maintaining accuracy by downsizing from a 3D CNN to a 2D CNN. As a result, we succeeded in increasing the processing speed by approximately 8 times while maintaining a certain degree of accuracy in inference.},
 title = {ドライバーの眠気予測を目的とした顔表情評定AIシステムの軽量化},
 year = {2022}
}