AgentPose: Progressive Distribution Alignment via Feature Agent for Human Pose Distillation

TL;DR

AgentPose introduces a feature agent-based progressive distribution alignment technique for human pose distillation, effectively bridging the capacity gap between teacher and student models to improve knowledge transfer.

Contribution

The paper presents a novel pose distillation method that models and progressively aligns feature distributions to overcome capacity gaps, enhancing transfer effectiveness.

Findings

Effective knowledge transfer on COCO dataset

Improved performance with large capacity gaps

Outperforms existing pose distillation methods

Abstract

Pose distillation is widely adopted to reduce model size in human pose estimation. However, existing methods primarily emphasize the transfer of teacher knowledge while often neglecting the performance degradation resulted from the curse of capacity gap between teacher and student. To address this issue, we propose AgentPose, a novel pose distillation method that integrates a feature agent to model the distribution of teacher features and progressively aligns the distribution of student features with that of the teacher feature, effectively overcoming the capacity gap and enhancing the ability of knowledge transfer. Our comprehensive experiments conducted on the COCO dataset substantiate the effectiveness of our method in knowledge transfer, particularly in scenarios with a high capacity gap.