Compressed Deep Convolution Neural Network for Face Recognition

DOI

10.2991/icmmita-16.2016.22How to use a DOI?

Keywords

Face recognition; Convolutional Neural Network; model compression.

Abstract

Deep convolution neural network (CNN) has achieved a great success on face recognition techniques. But most of CNN models tend to be much deeper, which are at the expenses of high consumption of computation and storage. So, it is hard for these deep CNNs applied to mobile equipments because of poor computational and memory resources. To alleviate this issue, this paper optimizes a lightened baseline CNN model by adopting an additional contrastive loss to learn more discriminative features. To further reduce the number of parameters, a pruning strategy is tried to compress our model, which slightly improves accuracy on the LFW dataset with the compression ratio of 0.7. Finally, experimental result shows that the proposed method achieve state-of-the-art results with much smaller size and fewer training data.

Copyright

© 2017, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Ying Zou
AU  - Xiaohong Liu
PY  - 2017/01
DA  - 2017/01
TI  - Compressed Deep Convolution Neural Network for Face Recognition
BT  - Proceedings of the 2016 4th International Conference on Machinery, Materials and Information Technology Applications
PB  - Atlantis Press
SP  - 110
EP  - 114
SN  - 2352-538X
UR  - https://doi.org/10.2991/icmmita-16.2016.22
DO  - 10.2991/icmmita-16.2016.22
ID  - Zou2017/01
ER  -