Seq-Masks: Bridging the gap between appearance and gait modeling for video-based person re-identification

ideo-based person re-identification (Re-ID) aims to match person images in\nvideo sequences captured by disjoint surveillance cameras. Traditional\nvideo-based person Re-ID methods focus on exploring appearance information,\nthus, vulnerable against illumination changes, scene noises, camera parameters,\nand especially clothes/carrying variations. Gait recognition provides an\nimplicit biometric solution to alleviate the above headache. Nonetheless, it\nexperiences severe performance degeneration as camera view varies. In an\nattempt to address these problems, in this paper, we propose a framework that\nutilizes the sequence masks (SeqMasks) in the video to integrate appearance\ninformation and gait modeling in a close fashion. Specifically, to sufficiently\nvalidate the effectiveness of our method, we build a novel dataset named\nMaskMARS based on MARS. Comprehensive experiments on our proposed large wild\nvideo Re-ID dataset MaskMARS evidenced our extraordinary performance and\ngeneralization capability. Validations on the gait recognition metric CASIA-B\ndataset further demonstrated the capability of our hybrid model.\n