多任务学习及卷积神经网络在人脸识别中的应用

随着深度学习的发展，近年来人脸识别借助深度学习技术取得了巨大突破。但是在已有的基于深度学习的人脸识别框架中，各个任务（人脸鉴别、认证和属性分类等）都是相互独立设计、运作的，使得整体算法低效、耗时。针对这些问题，提出一种基于多任务框架的深度卷积网络。通过将人脸鉴别、认证和属性分类同时作为网络目标函数，端到端地训练整个深度卷积网络，算法简洁高效。此网络可以同时完成上述三个任务，不需要额外的步骤。实验结果显示，即使在有限的数据支持下，该方法依然能够取得不错的性能，在人脸识别权威数据集LFW上获得了97.3%的精度。

Multitask learning and CNN for application of face recognition

With the development of deep learning, face recognition algorithm has made tremendous breakthroughs. However, among current face recognition frameworks, each task （face identification, face verification or attribute classification） is independently designed and manipulated, which makes the algorithm inefficient and time-consuming. According to the problem, this paper proposes a multi-task convolution deep network. By combining face identification, verification and attribute classification losses as this loss function, the deep convolution network can be trained from end to end and the algorithm will be simple and efficient. This network can complete these three tasks without additional steps. Experiments show that the model can still achieve good performance with limited training data and get 97.3% accuracy in the authoritative face recognition dataset LFW（Labeled Face in the Wild）.