Heatmap Regression without Soft-Argmax for Facial Landmark Detection

TL;DR

This paper introduces a new training approach for facial landmark detection that avoids the traditional Soft-argmax method, achieving faster training and competitive accuracy on multiple benchmarks.

Contribution

It proposes an alternative to Soft-argmax for heatmap regression, improving training speed and maintaining high accuracy in facial landmark detection.

Findings

Achieves state-of-the-art results on WFLW, COFW, and 300W datasets.

Converges 2.2x faster during training.

Maintains or improves accuracy compared to Soft-argmax-based methods.

Abstract

Facial landmark detection is an important task in computer vision with numerous applications, such as head pose estimation, expression analysis, face swapping, etc. Heatmap regression-based methods have been widely used to achieve state-of-the-art results in this task. These methods involve computing the argmax over the heatmaps to predict a landmark. Since argmax is not differentiable, these methods use a differentiable approximation, Soft-argmax, to enable end-to-end training on deep-nets. In this work, we revisit this long-standing choice of using Soft-argmax and demonstrate that it is not the only way to achieve strong performance. Instead, we propose an alternative training objective based on the classic structured prediction framework. Empirically, our method achieves state-of-the-art performance on three facial landmark benchmarks (WFLW, COFW, and 300W), converging 2.2x faster during training while maintaining better/competitive accuracy. Our code is available here: https://github.com/ca-joe-yang/regression-without-softarg.