UntrimmedNets for Weakly Supervised Action Recognition and Detection

Current action recognition methods heavily rely on trimmed videos for model training. However, it is very expensive and time-consuming to acquire a large-scale trimmed video dataset. This paper presents a new weakly supervised architecture, called UntrimmedNet, which is able to directly learn from untrimmed videos without the need of temporal annotations of action instances. Our UntrimmedNet couples two important components, the classification module and the selection module, to learn the action models and reason about the temporal duration of action instances, respectively. These two modules are implemented with feed-forward networks. UntrimmedNet is essentially an end-to-end trainable architecture, which allows for the joint optimization of model parameters of both components. We exploit the learned models for the problems of action recognition (WSR) and detection (WSD) on the untrimmed video datasets of THUMOS14 and ActivityNet. Although our UntrimmedNet only employs weak supervision, our method achieves performance superior or comparable to that of strongly supervised approaches on these two datasets.