论文部分内容阅读
In the field of crossing-scene pedestrian identification, the recognition accuracy is low due to the large local variation of the samples. A method based on twice Feature-Aggregation-Separation (FAS) is proposed in this paper. Firstly, a novel network str