Again-Pose: Anchor-Guided Adaptive Inter-Frame Motion Cues Propagating for High-quality Human Pose Reconstruction

2026-06-28

AI Summary

This work addresses the challenge of robust 3D human pose estimation under extreme degradations such as severe motion blur and occlusion, where existing methods suffer from feature collapse and ineffective temporal aggregation. To overcome these limitations, the authors reformulate pose estimation in degraded frames as a motion-guided recovery task and introduce an anchor-guided adaptive motion cue propagation mechanism. By identifying high-quality anchor frames, the method explicitly propagates their reliable kinematic information to reconstruct intermediate degraded frames, avoiding indiscriminate smoothing. A dual-path motion-aware module captures fine-grained dynamics, while a differential weighting fusion strategy adaptively suppresses pose drift. Evaluated on Human3.6M, 3DPW, PoseTrack, and FineDiving benchmarks, the proposed approach significantly outperforms state-of-the-art methods, consistently recovering plausible and stable poses even under extreme degradation conditions.

Abstract

Reconstructing continuous 3D human poses from unconstrained videos is challenging, especially in extreme motion scenarios involving severe motion blur and occlusion. Current state-of-the-art methods typically rely on implicit temporal attention to aggregate features across frames. However, under severe visual degradation, input features often suffer from collapse, rendering them indistinguishable from noise. In such cases, implicit aggregation fails to distinguish valid signals, leading to catastrophic reconstruction errors. To address this robustness gap, we propose a simple yet effective framework called Anchor-guided adaptive inter-frame motion cues propagating (Again-Pose), reformulating pose estimation in degraded frames as a motion-guided recovery task. Instead of blindly smoothing features, we explicitly identify high-quality Anchor Frames based on feature saliency and propagate reliable kinematic cues to "inpaint" the poses of degraded intermediate frames. Specifically, a Dual-path Motion-aware Module captures fine-grained inter-frame dynamics, while a Difference-weighted Fusion Module adaptively propagates these cues to suppress drift. Extensive experiments on standard benchmarks (Human3.6M, 3DPW, PoseTrack) and the challenging FineDiving dataset demonstrate that Again-Pose significantly outperforms state-of-the-art methods in robustness and stability, effectively recovering plausible poses where other methods fail.

Problem

• 3D human pose reconstruction
• motion blur
• occlusion
• temporal feature aggregation
• visual degradation

Innovation

• Anchor-guided
• motion cues propagation
• 3D human pose reconstruction
• feature saliency
• temporal robustness