CorrLoss: Integrating Co-Occurrence Domain Knowledge for Affect Recognition

TL;DR

This paper introduces CorrLoss, a novel loss function that incorporates co-occurrence domain knowledge of facial movements to improve neural network affect recognition, enhancing generalizability and reducing overfitting.

Contribution

The paper presents CorrLoss, a new method that embeds co-occurrence domain knowledge into neural network training for affect recognition, improving cross-dataset performance.

Findings

Enhanced cross-dataset testing accuracy

Improved model calibration for facial expressions

Reduced overfitting in affect recognition models

Abstract

Neural networks are widely adopted, yet the integration of domain knowledge is still underutilized. We propose to integrate domain knowledge about co-occurring facial movements as a constraint in the loss function to enhance the training of neural networks for affect recognition. As the co-ccurrence patterns tend to be similar across datasets, applying our method can lead to a higher generalizability of models and a lower risk of overfitting. We demonstrate this by showing performance increases in cross-dataset testing for various datasets. We also show the applicability of our method for calibrating neural networks to different facial expressions.