Ear detection using fully convolutional networks
Sida Wang, Yajun Du, Zengxi Huang · 2017
Automatic human ear detection has aroused great interest in biometrics community recently. Ear recognition has great potential in security application, especially it is a naturally complement to face recognition to strength identification recognition. Recent studies have shown that deep learning has achieved very good result in terms of object detection. In this article, we have proposed a practical method for human ear detection called FCNED. Firstly, we construct a human ear classifier based on convolutional neural network, and then transform it into a fully convolutional neural network. Finally, we utilize the sliding-window characteristic of the fully convolutional neural network for human ear detection. In order to improve the ear detection accuracy, the methods of multi-scale and NMS(non-maximum suppression) are also used in our paper. The results of experiment show that our method achieves a very good performance.