Graph-Attention Network for Cross-Domain Facial Expression Recognition
Deep Learning
Graph Neural Networks
Computer Vision
ResNet-50 and Graph Attention Networks with adversarial domain alignment, achieving 74% mean accuracy across six datasets.
Integrated ResNet-50 and Graph Attention Networks with adversarial domain alignment (MMD, CORAL, GRL). Achieved 74% mean accuracy across six datasets: RAF-DB, FER2013, CK+, JAFFE, ExpW, and SFEW2.0.
This work was presented at ACML 2025.