Fast video facial expression recognition by deeply tensor-compressed lstm neural network on mobile device

Zhen, Peining1; Liu, Bin2; Cheng, Yuan1; Chen, Hai-Bao1; Yu, Hao2

10.1145/3318216.3363322

2019

4th ACM/IEEE Symposium on Edge Computing, SEC 2019

298-300

Arlington, VA, United states

Association for Computing Machinery, Inc

Poster: Mobile devices usually suffer from limited computation and storage resource which seriously hinders them from deep neural network applications. In this paper, we introduce a deeply tensor-compressed LSTM neural network for fast facial expression recognition (FER) in videos on mobile devices. Firstly, a spatio-temporal FER LSTM model is built by extracting time-series feature maps from facial clips. The LSTM model is further deeply compressed with tensorization. Based on dataset of Acted Facial Expression in Wild (AFEW) 7.0, experimental results show that the proposed method achieves 55.60% classification accuracy; and significantly compresses the size of network model by 219×. Our work is further implemented on RK3399Pro IoT device with Neural Process Engine, and the runtime of feature extraction part can be reduced by 12.83× with only 7.73W power consumption.
© 2019 Copyright held by the owner/author(s).

其他

EI

20195007834055

Classification (of information) ; Deep learning ; Deep neural networks ; Edge computing ; Face recognition ; Mobile computing ; Tensors

Telecommunication; Radar, Radio and Television:716 ; Information Theory and Signal Processing:716.1 ; Algebra:921.1

EV Compendex

1.Shanghai Jiao Tong University, Shanghai, China
2.Southern University of Science and Technology, Shenzhen, China

Zhen, Peining,Liu, Bin,Cheng, Yuan,et al. Fast video facial expression recognition by deeply tensor-compressed lstm neural network on mobile device[C]:Association for Computing Machinery, Inc,2019:298-300.