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

TL;DR

This paper presents an end-to-end recurrent neural network approach for action detection in videos, where the model learns to observe and refine predictions over time using reinforcement learning, achieving state-of-the-art results with minimal frame observations.

Contribution

It introduces a novel agent-based model that directly predicts action boundaries in videos by learning to selectively observe frames, using reinforcement learning for training.

Findings

Achieves state-of-the-art results on THUMOS'14 and ActivityNet datasets.

Observes only 2% or less of video frames during detection.

Demonstrates effective action detection with minimal frame observations.

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.