Recently, the paper of Prof. Wei Zhang's group 'Reconstruct and Represent Video Contents for Captioning via Reinforcement Learning' was accepted as a full paper by the international top journal 'IEEE Transactions on Pattern Analysis and Machine Intelligence' (TPAMI). TPAMI is the top journal in the area of computer vision and pattern recognition, and is also among the class-A academic recommended list of international periodicals. The impact factor of TPAMI is 9.445.

This paper proposed a reconstruction network (RecNet) in a novel encoder-decoder-reconstructor architecture, which leverages both forward (video to sentence) and backward (sentence to video) flows for video captioning. The proposed model could reproduce the video features from local and global perspectives, which for the first time enables the two-way mappings between video and sentence. Experiments demonstrate that the RecNet model achieve sstate-of-the-art performances on several international benchmark datasets.

The lead author of this paper is a graduate student of Shandong University, Bairui Wang, supervised by Prof. Wei Zhang. This work was done in cooperation with Tencent AI Lab, which employs over 70 top scientists from universities with world-wide reputations, and cooperates with top universities and institutions for developing both practical and advanced AI products. The research products of Tencent AI Lab have been successfully applied in hundreds of Tencent products, including WeChat, QQ and KuaiBao. The group of Prof. Wei Zhang has a long term cooperation with TenCent AI Lab, and this work is a demonstration of their successful explorations of work-study combination.