Understanding and Mitigating Annotation Bias in Facial Expression Recognition

Yunliang Chen, Jungseock Joo

The performance of a computer vision model depends on the size and quality of its training data. Recent studies have unveiled previously-unknown composition biases in common image…