基于深度迁移学习的人脸识别方法研究

Study on Face Recognition Method Based on Deep Transfer Learning

针对大数据集上学习的深度人脸模型在实践中的相关问题,提出一种通过迁移一个预训练的深度人脸模型到特定的任务来解决该问题的方案:将深度人脸模型学习的分层表示作为源模型,然后在一个小训练集上学习高层表示以得到一个特定于任务的目标模型;在公共的小数据集及采集的真实人脸数据集上的实验表明,所采用的迁移学习方法有效且高效;经验性地探索了一个重要的开放问题——深度模型不同层特征的特点及其可迁移能力,认为越底层的特征越局部、越通用,而越高层的特征则越全局、越特定,具有更好的类内不变性和类间区分性;无监督的特征可视化与有监督的人脸识别实验结果都能较好地支持上述观点.

Aiming at relevant problems of deep face model for learning based on big dataset in practice, we put forward a scheme to deal with these problems through transferring a pre-training deep face model to specific tasks on hand. We empirically transfer hierarchical representations of deep face model as a source model and then learn higher layer representations on a specific small training set to obtain a final task- spe-cific target model. Experiments on face identification tasks with public small data set and practical real face data set verify the effectiveness and efficiency of our approach for transfer learning. We also empirically ex-plore an important open problem -attributes and transferability of different layer features of deep model. We argue that lower layer features are both local and general, while higher layer ones are both global and spe-cific which embraces both intra-class invariance and inter-class discrimination. The results of unsupervised feature visualization and supervised face identification strongly support our view.