Person re-identification (re-ID) remains challenging in a real-world\nscenario, as it requires a trained network to generalise to totally unseen\ntarget data in the presence of variations across domains. Recently, generative\nadversarial models have been widely adopted to enhance the diversity of\ntraining data. These approaches, however, often fail to generalise to other\ndomains, as existing generative person re-identification models have a\ndisconnect between the generative component and the discriminative feature\nlearning stage. To address the on-going challenges regarding model\ngeneralisation, we propose an end-to-end domain adaptive attention network to\njointly translate images between domains and learn discriminative re-id\nfeatures in a single framework. To address the domain gap challenge, we\nintroduce an attention module for image translation from source to target\ndomains without affecting the identity of a person. More specifically,\nattention is directed to the background instead of the entire image of the\nperson, ensuring identifying characteristics of the subject are preserved. The\nproposed joint learning network results in a significant performance\nimprovement over state-of-the-art methods on several benchmark datasets.\n