A Face Recognition System using Convolutional Neural Network and Generalized with Facial Expression

  IJCTT-book-cover
 
International Journal of Computer Trends and Technology (IJCTT)          
 
© 2017 by IJCTT Journal
Volume-49 Number-3
Year of Publication : 2017
Authors : C.R Vimalchand, Dr. G.P Ramesh Kumar
DOI :  10.14445/22312803/IJCTT-V49P126

MLA

C.R Vimalchand, Dr. G.P Ramesh Kumar "A Face Recognition System using Convolutional Neural Network and Generalized with Facial Expression". International Journal of Computer Trends and Technology (IJCTT) V49(3):161-168, July 2017. ISSN:2231-2803. www.ijcttjournal.org. Published by Seventh Sense Research Group.

Abstract -
Facial expression recognition systems have attracted much research interest within the field of artificial intel-ligence. Face Recognition can be applied for many security issues as a prompt solution. Old facial expression recognition (FER) systems apply standard machine learning to ex-tracted image features like geometric features and these methods generalize poorly to previously recorded database. This work introduces some re-cent research to classify images of human faces into dis-crete emotion categories using convolutional neural net-works (CNNs). We experimented with different architec-tures and methods such as fractional max-pooling and fine-tuning, ultimately achieving an accuracy of 0.48 in a seven-class classification task. The objective of this project is to classify images of hu-man faces into discrete emotion categories. Most of the established facial expression recognition (FER) systems use standard machine learning and extracted features, which do not have significant performance when applied to previously recorded database [1]. Within the past few months a few papers have been published that use deep learning for FER [2] [13] which have been successful at achieving about .60 accuracy on the EmotiW and other publicly available data sets. Not-ing the success of CNNs in this domain, our objective is to experiment with both new and existing network architec-tures to achieve similar results on a new data set.

References
[1] Ekman and W. V. Friesen. Emotional facial action coding system. Unpublished manuscript, University of California at San Francisco, 1983.
[2] B. Graham. Fractional max-pooling. 2015.
[3] S. Ioffe and C. Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. JMLR Proceedings, 2015.
[4] M. S.-S. S. M. B. Z. L. X. S. B. P. J. Zhang, S. Ma and Mech. Salient object subitizing. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015.
[5] Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Gir-shick, S. Guadarrama, and T. Darrell. Caffe: Convolu-tional architecture for fast feature embedding. arXiv preprint arXiv:1408.5093, 2014.
[6] Lab41. Misc, caffekeras util. https://github.com/Lab41/Misc/blob/master/blog/kerasvgg/ caffekeras_util.py, 2015.
[7] Lasange. Lasange, fractional pooling layer pull re-quest. https://github.com/Lasagne/Lasagne/ pull/171, 2015.
O. H. S.-J. K. S. S. S. M. Z. H. A. K. A. K. M. B. A. C. B.Russakovsky, J. Deng and L. Fei-Fei. Imagenet large scale visual recognition challenge. IJCV, 2015.
[8] A. V. Omkar M. Parkhi and A. Zisserman. Deep face recog-nition. 2015.
[9] C. W. Pablo Barros and S. Wermter. Emotional expression recognition with a cross-channel convolutional neural net-work for human-robot interaction. IEEE 15th International Conference on Humanoid Robots, 2015.
[10] Z. Yu and C. Zhang. Image based static facial expression recognition with multiple deep network learning. ICMI Pro-ceedings.
[12] D. C. Ali Mollahosseini and M. H. Mahoor. Going deeper in facial expression recognition using deep neural networks.
[13] IEEE Winter Conference on Applications of Computer Vi-sion, 2016.
[14] S.-Y. D. Bo-Kyeong Kim, Jihyeon Roh and S.-Y. Lee. Hi-erarchical committee of deep convolutional neural networks for robust facial expression recognition. Journal on Multi-modal User Interfaces, pages 1–17, 2015.

Keywords
experiment with both new and existing network architec-tures to achieve similar results on a new data set.