有信念，才能走的更远

1、We focus on blurred images from a real-life traffic surveillance system, on which we, for the first time, demonstrate that neural networks trained on artificial data provide superior reconstruction

1、In this paper, we address this problem by jointly learning a deep model for two tasks, i.e. face hallucination and recognition.本文是连接两个任务模型：hallucination和recognition。2、The recognition subnetwork

1、However, conventional hallucination methods are often designed for controlled settings and cannot handle varying conditions of pose, resolution degree, and blur.传统的方法有约束控制，不能处理多种姿态、像素深度及遮挡的情况。2、