说明：  在人脸检测和特征点定位的任务上，这篇文章提出的方法比现有最先进的技术有明显的额提升，而且具有实时处理的性能。 在这篇文章中，提出了一个深度卷积多任务的框架，这个框架利用了检测和对准之间固有的关系来增强他们的性能。特别是，在预测人脸及脸部标记点的时候，通过三阶的级联卷积神经网络对任务进行从粗到精的处理。而且还提出了一种新的在线困难样本生成策略来进一步提升性能。 论文最主要的一点应该是三阶的级联卷积神经网络。
(In the task of face detection and feature point location, the method proposed in this paper has a significant improvement compared with the most advanced technology, and has real-time processing performance. In this paper, a deep convolution multitask framework is proposed, which uses the inherent relationship between detection and alignment to enhance their performance. Especially, in the prediction of face and facial markers, the tasks are processed from coarse to fine by the third-order cascaded convolutional neural network. In addition, a new online difficult sample generation strategy is proposed to further improve the performance. The main point of this paper is the third-order cascaded convolutional neural network.)