TURN TAP: Temporal Unit Regression Network for Temporal Action Proposals

TLDR

Temporal Action Proposal generation is crucial for efficiently extracting human action segments from untrimmed videos in large‑scale video analysis. The authors introduce the Temporal Unit Regression Network (TURN) to generate temporal action proposals. TURN jointly predicts proposals and refines their temporal boundaries via coordinate regression while achieving fast computation by reusing video units as reusable building blocks. TURN surpasses prior state‑of‑the‑art methods in average recall on THUMOS‑14 and ActivityNet, processes videos at over 880 FPS on a TITAN X GPU, and further boosts localization accuracy when integrated into existing pipelines on both datasets.

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 previous 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.