Meta-Learned Adaptive Optimization for Robust Human Mesh Recovery with Uncertainty-Aware Parameter Updates

2026-03-27

AI Summary

This work addresses the challenges of single-image human mesh reconstruction, which is often hindered by depth ambiguity and limited cross-domain generalization, as well as the inefficacy of existing test-time optimization methods due to poor initialization and suboptimal parameter updates. To overcome these limitations, the authors propose a meta-learning-driven adaptive optimization framework that learns optimization-friendly initial parameters during meta-training. The approach incorporates a selective parameter caching mechanism that freezes converged joints and introduces an uncertainty-aware update strategy based on distribution sampling, enabling efficient and robust test-time refinement. Evaluated on 3DPW and Human3.6M, the method reduces MPJPE by 10.3 and 8.0, respectively, significantly outperforming strong baselines. Moreover, it produces uncertainty estimates highly correlated with prediction errors, demonstrating exceptional cross-domain adaptability.

Abstract

Human mesh recovery from single images remains challenging due to inherent depth ambiguity and limited generalization across domains. While recent methods combine regression and optimization approaches, they struggle with poor initialization for test-time refinement and inefficient parameter updates during optimization. We propose a novel meta-learning framework that trains models to produce optimization-friendly initializations while incorporating uncertainty-aware adaptive updates during test-time refinement. Our approach introduces three key innovations: (1) a meta-learning strategy that simulates test-time optimization during training to learn better parameter initializations, (2) a selective parameter caching mechanism that identifies and freezes converged joints to reduce computational overhead, and (3) distribution-based adaptive updates that sample parameter changes from learned distributions, enabling robust exploration while quantifying uncertainty. Additionally, we employ stochastic approximation techniques to handle intractable gradients in complex loss landscapes. Extensive experiments on standard benchmarks demonstrate that our method achieves state-of-the-art performance, reducing MPJPE by 10.3 on 3DPW and 8.0 on Human3.6M compared to strong baselines. Our approach shows superior domain adaptation capabilities with minimal performance degradation across different environmental conditions, while providing meaningful uncertainty estimates that correlate with actual prediction errors. Combining meta-learning and adaptive optimization enables accurate mesh recovery and robust generalization to challenging scenarios.

Problem

• human mesh recovery
• depth ambiguity
• domain generalization
• test-time optimization
• parameter initialization

Innovation

• meta-learning
• adaptive optimization
• uncertainty-aware updates
• parameter initialization
• human mesh recovery