Pose-driven attention-guided image generation for person re-Identification

Person re-identification (re-ID) concerns the matching of subject images across different camera views in a multi surveillance system. One major challenges person re-ID is pose variations network, which significantly affects appearance person. Existing development data lack adequate to carry out effective training systems. To solve this issue, paper we propose an end-to-end pose-driven attention-guided generative adversarial generate multiple poses We attentively learn and transfer through attention mechanism. A semantic-consistency loss proposed preserve semantic information during transfer. ensure fine image details are realistic after translation, discriminator used while transferred will exactly be same as target pose. show that by incorporating approach framework, state-of-the-art results can achieved.