What have we learned from deep representations for action recognition?
This work provides insights into how deep models capture actions, which is important for researchers in computer vision seeking to understand and improve video recognition systems, though it is incremental as it builds on existing two-stream models.
The paper investigates deep spatiotemporal representations for action recognition in video by visualizing two-stream models, revealing that cross-stream fusion enables true spatiotemporal features and networks learn both class-specific and generic representations.
As the success of deep models has led to their deployment in all areas of computer vision, it is increasingly important to understand how these representations work and what they are capturing. In this paper, we shed light on deep spatiotemporal representations by visualizing what two-stream models have learned in order to recognize actions in video. We show that local detectors for appearance and motion objects arise to form distributed representations for recognizing human actions. Key observations include the following. First, cross-stream fusion enables the learning of true spatiotemporal features rather than simply separate appearance and motion features. Second, the networks can learn local representations that are highly class specific, but also generic representations that can serve a range of classes. Third, throughout the hierarchy of the network, features become more abstract and show increasing invariance to aspects of the data that are unimportant to desired distinctions (e.g. motion patterns across various speeds). Fourth, visualizations can be used not only to shed light on learned representations, but also to reveal idiosyncracies of training data and to explain failure cases of the system.