| 摘 要: 针对分层人脸年龄估计方法在年龄阶段边界样本上预测不稳定的问题,本文提出一种基于年龄阶段不确定性感知的分层人脸年龄估计方法。该方法首先通过阶段概率分布刻画样本在不同年龄阶段上的语义关系,并利用信息熵量化阶段不确定性,从而获得连续的阶段语义表示。在此基础上,构建不确定性引导的分层决策机制,将阶段不确定性映射为分层调制系数,使模型能够根据阶段语义的可靠程度自适应调节阶段约束在预测过程中的作用强度。进一步地,在回归阶段设计不确定性调制的分层回归融合框架,对阶段约束回归与全局回归进行动态融合,从而提升模型在复杂样本上的预测稳定性。实验结果表明,该方法在MORPH II、AFAD和CACD等公开数据集上取得了稳定且具有竞争力的性能。 |
| 关键词: 分层人脸年龄估计 年龄阶段不确定性 分层决策机制 |
|
中图分类号:
文献标识码:
|
|
| The Uncertainty-Aware Hierarchical Facial Age Estimation Method |
|
zhangjiarong, huchunlong
|
Jiangsu University of Science and Technology
|
| Abstract: To address the instability of hierarchical facial age estimation methods on samples located near age-stage boundaries, this paper proposes an uncertainty-aware hierarchical facial age estimation method based on age-stage uncertainty perception. The proposed method first models the semantic relationships of samples across different age stages using a stage probability distribution, and quantifies stage uncertainty through information entropy to obtain a continuous representation of stage semantics. On this basis, an uncertainty-guided hierarchical decision mechanism is constructed, where stage uncertainty is mapped into a hierarchical modulation coefficient, enabling the model to adaptively adjust the influence of stage constraints during prediction according to the reliability of stage semantics. Furthermore, an uncertainty-modulated hierarchical regression fusion framework is designed in the regression stage to dynamically integrate stage-constrained regression and global regression, thereby improving prediction stability on complex samples. Experimental results demonstrate that the proposed method achieves stable and competitive performance on several public datasets, including MORPH II, AFAD, and CACD. |
| Keywords: Hierarchical facial age estimation Age-stage uncertainty perception hierarchical decision mechanism |
