LAE : Long-tailed Age Estimation

TL;DR

This paper introduces LAE, a two-stage training method for facial age estimation that addresses data imbalance and improves accuracy by decoupling representation learning from classification.

Contribution

It proposes a novel two-stage training framework for long-tailed age estimation, enhancing performance on imbalanced datasets.

Findings

Significant reduction in age estimation errors compared to baselines.

Effective handling of long-tailed age data, especially for elderly and children.

Validated on Guess The Age Contest 2021 dataset.

Abstract

Facial age estimation is an important yet very challenging problem in computer vision. To improve the performance of facial age estimation, we first formulate a simple standard baseline and build a much strong one by collecting the tricks in pre-training, data augmentation, model architecture, and so on. Compared with the standard baseline, the proposed one significantly decreases the estimation errors. Moreover, long-tailed recognition has been an important topic in facial age datasets, where the samples often lack on the elderly and children. To train a balanced age estimator, we propose a two-stage training method named Long-tailed Age Estimation (LAE), which decouples the learning procedure into representation learning and classification. The effectiveness of our approach has been demonstrated on the dataset provided by organizers of Guess The Age Contest 2021.