End-to-End Learning of Action Detection from Frame Glimpses in Videos

TLDR

The action detection process can be viewed as observation and refinement, where moments in video are observed and hypotheses about action timing are iteratively refined. This work introduces a fully end‑to‑end approach that directly predicts temporal action bounds in videos. We formulate a recurrent neural network–based agent that interacts with a video over time, observing frames to decide where to look next and when to emit a prediction, and we train its decision policy with REINFORCE. The model achieves state‑of‑the‑art results on THUMOS'14 and ActivityNet while observing only 2 % or fewer of the video frames.

Abstract

In this work we introduce a fully end-to-end approach for action detection in videos that learns to directly predict the temporal bounds of actions. Our intuition is that the process of detecting actions is naturally one of observation and refinement: observing moments in video, and refining hypotheses about when an action is occurring. Based on this insight, we formulate our model as a recurrent neural network-based agent that interacts with a video over time. The agent observes video frames and decides both where to look next and when to emit a prediction. Since backpropagation is not adequate in this non-differentiable setting, we use REINFORCE to learn the agent's decision policy. Our model achieves state-of-the-art results on the THUMOS'14 and ActivityNet datasets while observing only a fraction (2% or less) of the video frames.