Action Search: Spotting Actions in Videos and Its Application to Temporal Action Localization

TL;DR

This paper introduces Action Search, a neural network model inspired by human action spotting behavior, which efficiently locates actions in videos by observing only a small portion, and provides a new dataset of human search sequences.

Contribution

It proposes a novel action spotting problem, a human-inspired RNN approach, and the Human Searches dataset for training and evaluation.

Findings

Explores only 17.3% of videos on average

Achieves 30.8% mAP in temporal action localization

Demonstrates efficient and accurate action spotting

Abstract

State-of-the-art temporal action detectors inefficiently search the entire video for specific actions. Despite the encouraging progress these methods achieve, it is crucial to design automated approaches that only explore parts of the video which are the most relevant to the actions being searched for. To address this need, we propose the new problem of action spotting in video, which we define as finding a specific action in a video while observing a small portion of that video. Inspired by the observation that humans are extremely efficient and accurate in spotting and finding action instances in video, we propose Action Search, a novel Recurrent Neural Network approach that mimics the way humans spot actions. Moreover, to address the absence of data recording the behavior of human annotators, we put forward the Human Searches dataset, which compiles the search sequences employed by human annotators spotting actions in the AVA and THUMOS14 datasets. We consider temporal action localization as an application of the action spotting problem. Experiments on the THUMOS14 dataset reveal that our model is not only able to explore the video efficiently (observing on average 17.3% of the video) but it also accurately finds human activities with 30.8% mAP.