TURN TAP: Temporal Unit Regression Network for Temporal Action Proposals

TL;DR

TURN is a fast and accurate temporal action proposal network that jointly predicts and refines action segments in untrimmed videos, significantly improving over previous methods in both accuracy and speed.

Contribution

The paper introduces TURN, a novel network that combines proposal prediction and boundary refinement with efficient unit feature reuse for superior temporal action proposals.

Findings

Outperforms state-of-the-art methods in average recall on THUMOS-14 and ActivityNet.

Achieves over 880 FPS on a TITAN X GPU.

Enhances existing action localization pipelines with better proposal generation.

Abstract

Temporal Action Proposal (TAP) generation is an important problem, as fast and accurate extraction of semantically important (e.g. human actions) segments from untrimmed videos is an important step for large-scale video analysis. We propose a novel Temporal Unit Regression Network (TURN) model. There are two salient aspects of TURN: (1) TURN jointly predicts action proposals and refines the temporal boundaries by temporal coordinate regression; (2) Fast computation is enabled by unit feature reuse: a long untrimmed video is decomposed into video units, which are reused as basic building blocks of temporal proposals. TURN outperforms the state-of-the-art methods under average recall (AR) by a large margin on THUMOS-14 and ActivityNet datasets, and runs at over 880 frames per second (FPS) on a TITAN X GPU. We further apply TURN as a proposal generation stage for existing temporal action localization pipelines, it outperforms state-of-the-art performance on THUMOS-14 and ActivityNet.