Abstract: Convolutional network really necessary for learning a good visual representation for videobased person re-identification (VPRe-id)? In this paper, we first show that the common practice of employing convolutional neural networks (CNNs) to aggregate temporal spatial features may not be optimal. Specifically, with a diagnostic analysis, we show that the recurrent structure may not be effective learn temporal dependencies than what we expected and implicitly yields an order less representation. Based on this observation, we then present a simple yet surprisingly powerful approach for VPRe-id, where we treat VPRe-id as an efficient order less ensemble of image-based person re-identification problem. More specifically, we divide videos into individual images and re-identify person with ensemble of image-based rankers. Under the i.e. assumption, we provide an error bound that sheds light upon how could we improve VPRe-id. Our work also presents a promising way to bridge the gap between video and image-based person re-identification. Comprehensive experimental evaluations demonstrate that the proposed solution achieves state-of-the-art performances on multiple widely used various datasets.
Keyword: video, person re-identification, convolutional neural network.
Copyright Form