Two-Stream Convolutional Networks for Action Recognition in Videos
This work addresses the problem of video action recognition for computer vision applications, offering an incremental improvement by integrating spatial and temporal networks within a deep learning framework.
The paper tackles action recognition in videos by proposing a two-stream convolutional network architecture that separately processes spatial (appearance) and temporal (motion) information, achieving competitive state-of-the-art performance on UCF-101 and HMDB-51 benchmarks and significantly outperforming prior deep learning methods.
We investigate architectures of discriminatively trained deep Convolutional Networks (ConvNets) for action recognition in video. The challenge is to capture the complementary information on appearance from still frames and motion between frames. We also aim to generalise the best performing hand-crafted features within a data-driven learning framework. Our contribution is three-fold. First, we propose a two-stream ConvNet architecture which incorporates spatial and temporal networks. Second, we demonstrate that a ConvNet trained on multi-frame dense optical flow is able to achieve very good performance in spite of limited training data. Finally, we show that multi-task learning, applied to two different action classification datasets, can be used to increase the amount of training data and improve the performance on both. Our architecture is trained and evaluated on the standard video actions benchmarks of UCF-101 and HMDB-51, where it is competitive with the state of the art. It also exceeds by a large margin previous attempts to use deep nets for video classification.