L2A: Learning to Accumulate Pose History for Accurate 3D Human Pose Estimation
For researchers in 3D human pose estimation, this work introduces a novel method to effectively reuse cross-layer features, improving accuracy over existing approaches.
The paper tackles the problem of 2D-to-3D human pose estimation by addressing the underutilization of historical pose representations across network layers. The proposed history-aware framework with a spatial-temporal parallel Transformer and History Pose Accumulation mechanism achieves state-of-the-art performance on benchmarks.
Existing 2D-3D lifting human pose estimation methods have achieved strong performance. But the utilization of historical pose representations across network depth was overlooked. In current pipelines, information is propagated through fixed residual connections, which restricts effective reuse of early-layer features such as fine-grained spatial structures and short-term motion cues. However, naively incorporating historical features across layers is non-trivial. We further identify that maintaining a consistent representation space across layers is a prerequisite for effective cross-layer feature aggregation. To address this issue, we propose a history-aware framework that enables effective network cross-layer history feature utilization. Specifically, we adopt a spatial-temporal parallel Transformer backbone to prevent alternating spatial-temporal transformations during sequential processing, thereby maintaining a consistent representation space. Building upon this, we introduce a History Pose Accumulation (HPA) mechanism that adaptively aggregates features from all preceding layers to enhance current representations. Furthermore, we propose a Layer Pose History Aggregation (LPA) module that transforms layer pose features into a compact and structured form, reducing redundancy and enabling more stable aggregation. Extensive experiments demonstrate that our approach achieves state-of-the-art performance on benchmarks.