HSEmotion Team at ABAW-10 Competition: Facial Expression Recognition, Valence-Arousal Estimation, Action Unit Detection and Fine-Grained Violence Classification

HSEmotion Team developed a fast approach for facial emotion analysis using pre-trained EfficientNet models for the ABAW-10 competition. Their method combines confidence-based predictions with multi-layered perceptrons and sliding window smoothing, achieving significant improvements over existing baselines across four emotion recognition tasks.

• →HSEmotion Team presented results for the 10th Affective Behavior Analysis in-the-Wild competition focusing on facial emotion understanding.
• →The approach uses pre-trained EfficientNet-based models with confidence thresholds to determine prediction methods.
• →When model confidence is low, facial embeddings are processed through multi-layered perceptrons trained on AffWild2 dataset.
• →The system applies sliding window smoothing to reduce noise in frame-wise predictions.
• →Experimental results showed significant improvements over existing baselines across four ABAW challenge tasks.