Abstract | ||
---|---|---|
Over the past few years, Convolutional Neural Networks (CNNs) have shown promise on facial expression recognition. However, the performance degrades dramatically under real-world settings due to variations introduced by subtle facial appearance changes, head pose variations, illumination changes, and occlusions. In this paper, a novel island loss is proposed to enhance the discriminative power of deeply learned features. Specifically, the island loss is designed to reduce the intra-class variations while enlarging the inter-class differences simultaneously. Experimental results on four benchmark expression databases have demonstrated that the CNN with the proposed island loss (IL-CNN) outperforms the baseline CNN models with either traditional softmax loss or center loss and achieves comparable or better performance compared with the state-of-the-art methods for facial expression recognition. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/FG.2018.00051 | 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2018) |
Keywords | DocType | Volume |
Facial Expression Recognition,Island Loss,Convolutional Neural Network | Conference | abs/1710.03144 |
ISSN | ISBN | Citations |
2326-5396 | 978-1-5386-2336-7 | 17 |
PageRank | References | Authors |
0.60 | 0 | 6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Jie Cai | 1 | 57 | 4.77 |
Zibo Meng | 2 | 248 | 13.60 |
KHAN, AHMED-SHEHAB | 3 | 31 | 3.47 |
Zhiyuan Li | 4 | 30 | 8.40 |
James O'Reilly | 5 | 22 | 3.02 |
Yan Tong | 6 | 244 | 9.93 |