Multi-Level Network for High-Speed Multi-Person Pose Estimation

TL;DR

This paper introduces a Multi-level Network (MLN) for high-speed multi-person pose estimation that effectively discriminates joint types by aggregating multi-level features, achieving real-time performance.

Contribution

The proposed MLN efficiently combines features from multiple levels to improve joint type discrimination without increasing computational cost.

Findings

Achieves 42.2 FPS in multi-person pose estimation.

Performs comparably to traditional methods with less computation.

Effectively integrates multi-level features for joint discrimination.

Abstract

In multi-person pose estimation, the left/right joint type discrimination is always a hard problem because of the similar appearance. Traditionally, we solve this problem by stacking multiple refinement modules to increase network's receptive fields and capture more global context, which can also increase a great amount of computation. In this paper, we propose a Multi-level Network (MLN) that learns to aggregate features from lower-level (left/right information), upper-level (localization information), joint-limb level (complementary information) and global-level (context) information for discrimination of joint type. Through feature reuse and its intra-relation, MLN can attain comparable performance to other conventional methods while runtime speed retains at 42.2 FPS.