Facial Expression Recognition (FER) is crucial in various fields. This study focuses on enhancing FER through semi-supervised learning and temporal modeling. The limited FER dataset size hinders generalization, prompting the use of pseudo-labels for unlabeled data. A debiased feedback strategy addresses category imbalance and data bias. Introducing a Temporal Encoder captures temporal relationships for dynamic recognition. The method excelled in the 6th ABAW competition, confirming its effectiveness.
To Another Language
from source content
arxiv.org
Principais Insights Extraídos De
by Jun Yu,Zhiho... às arxiv.org 03-19-2024
https://arxiv.org/pdf/2403.11942.pdfPerguntas Mais Profundas