In the last decades, action recognition task has evolved from single view recording to unconstrained environment. Recently, multi-view action recognition has become a hot topic in computer vision. However, we… Click to show full abstract
In the last decades, action recognition task has evolved from single view recording to unconstrained environment. Recently, multi-view action recognition has become a hot topic in computer vision. However, we notice that only a few works have focused on the open-view action recognition, which is a common problem in the real world. Open-view action recognition focus on doing action recognition in unseen view without using any information from it. To address this issue, we firstly introduce a novel multi-view surveillance action dataset and benchmark several state-of-the-art algorithms. From the results, we observe that the performance of the state-of-the-art algorithms would drop a lot under open-view constraints. Then, we propose a novel open-view action recognition method based on the linear discriminant analysis. This method can learn a common space for action samples under different view by using their category information, which can achieve a good result in open-view action recognition.
               
Click one of the above tabs to view related content.