Attentional Group Convolution Neural Network for Light-weight Face Recognition on On-board Computer
Konferenz: MEMAT 2022 - 2nd International Conference on Mechanical Engineering, Intelligent Manufacturing and Automation Technology
07.01.2022 - 09.01.2022 in Guilin, China
Tagungsband: MEMAT 2022
Seiten: 6Sprache: EnglischTyp: PDF
Autoren:
Wang, Zhihao (School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou, Guangdong, China)
Jiang, Libiao (School of Mechanical Engineering, Guangzhou City University of Technology, Guangzhou, Guangdong, China)
Jiang, Siyu (Guangzhou Key Laboratory of Multilingual Intelligent Processing, School of Information Science and Technology, Guangdong University of Foreign Studies, Guangzhou, Guangdong, China)
Wang, Qin; Li, Yanming (Jiangling Motor Co., Ltd, Nanchang, Jiangxi, China)
Zhang, Weilin (Guangzhou Automobile Group Co., Ltd., Automobile Engineering Institute, Guangzhou, Guangdong, China)
Zheng, Chunyun (Automotive Engineering Research Institute, Guangzhou Automobile Group Co., Ltd, Guangzhou, Guangdong, China)
Inhalt:
Lightweight convolutional neural networks have been proposed to apply to hardware with poor performance for correct recognition of face identity, but still have the problems of large model and the low resolution of image sensors such as cameras leads to a sharp decline in recognition efficiency. A novel lightweight neural network model named Attentional Group Convolution Neural Network (AGCNet) with group convolution, channel shuffle and special attention modules is proposed for vehicle unlocking by face recognition on on-board computer. The experimental results indicate that compared with the existing models, the recognition accuracy of our model is significantly improved and the size of the model is significantly reduced, which achieves the state-of-the-art results with significant improvements.