Leaving Some Facial Features Behind

TL;DR

This paper investigates how removing or masking facial features affects emotion recognition accuracy, revealing that some features are critical while others may be less important or even beneficial to omit, leading to improved classification with a new training scheme.

Contribution

The study introduces a Perturb Scheme with three phases to enhance emotion classification by masking facial features during training, showing potential benefits of feature removal.

Findings

Removing key facial features reduces accuracy significantly for some emotions.

Masking features can unexpectedly improve accuracy for certain emotions like disgust.

The proposed Perturb Scheme improves classification performance through attention, clustering, and feature-focused training.

Abstract

Facial expressions are crucial to human communication, offering insights into emotional states. This study examines how specific facial features influence emotion classification, using facial perturbations on the Fer2013 dataset. As expected, models trained on data with the removal of some important facial feature experienced up to an 85% accuracy drop when compared to baseline for emotions like happy and surprise. Surprisingly, for the emotion disgust, there seem to be slight improvement in accuracy for classifier after mask have been applied. Building on top of this observation, we applied a training scheme to mask out facial features during training, motivating our proposed Perturb Scheme. This scheme, with three phases-attention-based classification, pixel clustering, and feature-focused training, demonstrates improvements in classification accuracy. The experimental results obtained suggests there are some benefits to removing individual facial features in emotion recognition tasks.