دانلود رایگان مقاله انگلیسی انطباق دامنه عمیق بررسی نشده برای تشخیص چهره - IEEE 2018

Abstract

Face recognition is challenge task which involves determining the identity of facial images. With availability of a massive amount of labeled facial images gathered from Internet, deep convolution neural networks(DCNNs) have achieved great success in face recognition tasks. Those images are gathered from unconstrain environment, which contain people with different ethnicity, age, gender and so on. However, in the actual application scenario, the target face database may be gathered under different conditions compered with source training dataset, e.g. different ethnicity, different age distribution, disparate shooting environment. These factors increase domain discrepancy between source training database and target application database and make the learnt model degenerate in target database. Meanwhile, for the target database where labeled data are lacking or unavailable, directly using target data to finetune pre-learnt model becomes intractable and impractical. In this paper, we adopt unsupervised transfer learning methods to address this issue. To alleviate the discrepancy between source and target face database and ensure the generalization ability of the model, we constrain the maximum mean discrepancy (MMD) between source database and target database and utilize the massive amount of labeled facial images of source database to training the deep neural network at the same time. We evaluate our method on two face recognition benchmarks and significantly enhance the performance without utilizing the target label.

V. CONCLUSIONS

In this paper, we focus on the issue of domain discrepancy between source training data and target test data in face recognition scenario. We adopt a deep unsupervised domain adaptation neural network and jointly utilize the labeled large scale source data and unlabeled target data to alleviate the domain discrepancy. We show the transferability between source face and target face by the multi-kernels MMD constraining on multi-layers representation. Empirical results show that the method can significantly enhance model performance on target test data without utilizing the label information.