Embarrassingly Simple Model for Early Action Proposal
This work addresses the need for efficient online action detection in video streams, though it appears incremental as it builds on existing methods with a simpler approach.
The paper tackled the problem of generating high-quality candidate temporal segments for actions in video streams as they happen, focusing on the online version of early action proposal. The result was a simple classifier-based model using standard 3D CNNs that performed significantly better than the state of the art.
Early action proposal consists in generating high quality candidate temporal segments that are likely to contain an action in a video stream, as soon as they happen. Many sophisticated approaches have been proposed for the action proposal problem but from the off-line perspective. On the contrary, we focus on the on-line version of the problem, proposing a simple classifier-based model, using standard 3D CNNs, that performs significantly better than the state of the art.