Description of Invention
Person re-identification (re-id) is the use of artificial intelligence techniques to determine the presence of a target person in an image or video sequence. In real-world scenarios, the pedestrian pose variations brings a formidable challenge to achieving high accuracy in recognition. To address this challenge, we design a novel dual-stream network (PGFNet), which introduces a local feature learning structure that comparatively learns pedestrian features with different poses, thereby enhancing the representation of pedestrians in various poses. PGFNet achieves impressive results on three highly competitive benchmarks and stands out as a superior solution.