Segregated Temporal Assembly Recurrent Networks for Weakly Supervised Multiple Action Detection
This addresses the problem of reducing annotation costs for action detection in videos, offering a weakly-supervised approach that achieves competitive results, though it is incremental as it builds on existing attention and recurrent methods.
The paper tackles weakly-supervised multiple action detection in untrimmed videos using only video-level labels, proposing a segregated temporal assembly recurrent (STAR) network that outperforms state-of-the-art weakly-supervised methods and matches fully-supervised performance on THUMOS'14 and ActivityNet1.3 datasets.
This paper proposes a segregated temporal assembly recurrent (STAR) network for weakly-supervised multiple action detection. The model learns from untrimmed videos with only supervision of video-level labels and makes prediction of intervals of multiple actions. Specifically, we first assemble video clips according to class labels by an attention mechanism that learns class-variable attention weights and thus helps the noise relieving from background or other actions. Secondly, we build temporal relationship between actions by feeding the assembled features into an enhanced recurrent neural network. Finally, we transform the output of recurrent neural network into the corresponding action distribution. In order to generate more precise temporal proposals, we design a score term called segregated temporal gradient-weighted class activation mapping (ST-GradCAM) fused with attention weights. Experiments on THUMOS'14 and ActivityNet1.3 datasets show that our approach outperforms the state-of-the-art weakly-supervised method, and performs at par with the fully-supervised counterparts.