AgentPose: Progressive Distribution Alignment via Feature Agent for Human Pose Distillation
This work addresses a specific bottleneck in model compression for human pose estimation, offering an incremental improvement over existing distillation methods.
The paper tackles the performance degradation in human pose distillation due to the capacity gap between teacher and student models by proposing AgentPose, which uses a feature agent to progressively align feature distributions, achieving improved knowledge transfer as validated on the COCO dataset.
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.