ZSTAD: Zero-Shot Temporal Activity Detection
This addresses the limitation of needing large annotated datasets for video surveillance and analysis, enabling detection of unseen activities, though it is an incremental advance in zero-shot learning for temporal detection.
The paper tackles the problem of detecting activities in videos without training examples for those activities, proposing a zero-shot temporal activity detection (ZSTAD) method that uses an end-to-end network with a novel loss function, achieving promising performance on THUMOS14 and Charades datasets.
An integral part of video analysis and surveillance is temporal activity detection, which means to simultaneously recognize and localize activities in long untrimmed videos. Currently, the most effective methods of temporal activity detection are based on deep learning, and they typically perform very well with large scale annotated videos for training. However, these methods are limited in real applications due to the unavailable videos about certain activity classes and the time-consuming data annotation. To solve this challenging problem, we propose a novel task setting called zero-shot temporal activity detection (ZSTAD), where activities that have never been seen in training can still be detected. We design an end-to-end deep network based on R-C3D as the architecture for this solution. The proposed network is optimized with an innovative loss function that considers the embeddings of activity labels and their super-classes while learning the common semantics of seen and unseen activities. Experiments on both the THUMOS14 and the Charades datasets show promising performance in terms of detecting unseen activities.