1. Introduction
Group re-identification (GReID) aims to match groups with the same members under different cameras. Usually, we deal with groups of 2 to 6 members, and we treat group images with more than 60% of the same members as the same group class. GReID aims to bring positive services and contributions to human society and eliminate poten-tial social risks, such as child trafficking and kidnapping. GReID has potential applications in detecting and preventing these events, which protects the safety of citizens. The challenge of GReID is how to jointly model the appearance and layout features of group images.
The illustration of our novelty. The X - y plane represents the imaging plane. The depth dimension represents the distance from person to camera. In the depth map, the darker the color, the closer to the camera, and vice versa.