A Hybrid 2D and 3D Convolution Based Recurrent Network for Video-Based Person Re-identification

Video-based person re-identification (re-id), which aims to match people through videos captured by non-overlapping camera views, has attracted lots of research interest recently. In this paper, we propose a novel hybrid 2D and 3D convolution based recurrent neural network for video-based person re-id task, which can simultaneously make use of the local short-term fast-varying motion information and the global long-term spatial and temporal information. Specifically, the 3D convolutional module is able to explore the local short-term fast-varying motion information, while the recurrent layer performed can learn global long-term spatial and temporal information. We evaluate the proposed hybrid neural network on the publicly available PRID 2011, iLIDS-VID and MARS multi-shot pedestrian re-identification datasets, and the experiment results demonstrate the effectiveness of our approach on the task of video-based person re-id.