Xu Dongshu

TL;DR: A Rotation Invariant Network to find the identical areas in cross-view images to extract robust local features in person re-identification images and shows the effectiveness of the method on public datasets including CUHK03, Market1501 and DukeMTMC.

TL;DR: In this paper, a depth feature representation method based on multiple stacked auto-encoding networks of different structures is proposed, where a shallow-layer neural network structure is constructed, a back-propagation method is adopted to train network parameters to enable a neural network to achieve the optimal structure, and outputs, namely the feature expressions, of a second layer of the network are acquired.

TL;DR: In this article, a pedestrian re-identification method combining global characteristics, part area characteristics and cross-view-anglesame area characteristics is constructed through the geometrical relationship between human body key points and the crossview-angle same area.

TL;DR: In this article, a specific figure and scene combined video instance retrieval method and system is presented, which comprises the following steps of: retrieving an instance of a particular figure in a video, retrieving a specific scene on the basis of local and global combinatorial optimization; retrieving a video instance on basis of high mark retaining; and fusing a specific Figure retrieval result and a specific Scene retrieval result, for each lens, fusing an initial Scene retrieval results and a Figure retrieval results after adjacent extension, and taking a maximum value of the two fusing results so as to obtain a