Hierarchical Attention-based Age Estimation and Bias Estimation

TL;DR

This paper introduces a hierarchical attention-based deep learning method for face age estimation that combines image augmentation, Transformer-Encoder aggregation, and probabilistic regression to improve accuracy and analyze biases.

Contribution

It presents a novel dual augmentation-aggregation approach with attention, and a probabilistic hierarchical regression framework for enhanced age estimation.

Findings

Outperforms existing age estimation methods on MORPH II dataset

Achieves state-of-the-art accuracy in face age estimation

Provides a bias analysis of current age estimation models

Abstract

In this work we propose a novel deep-learning approach for age estimation based on face images. We first introduce a dual image augmentation-aggregation approach based on attention. This allows the network to jointly utilize multiple face image augmentations whose embeddings are aggregated by a Transformer-Encoder. The resulting aggregated embedding is shown to better encode the face image attributes. We then propose a probabilistic hierarchical regression framework that combines a discrete probabilistic estimate of age labels, with a corresponding ensemble of regressors. Each regressor is particularly adapted and trained to refine the probabilistic estimate over a range of ages. Our scheme is shown to outperform contemporary schemes and provide a new state-of-the-art age estimation accuracy, when applied to the MORPH II dataset for age estimation. Last, we introduce a bias analysis of state-of-the-art age estimation results.