Multimodal depression severity estimation from clinical interviews — BiLSTM temporal attention on facial Action Units + MentalRoBERTa text embeddings, fused via SVM. PHQ-8 prediction on E-DAIC dataset.
python computer-vision deep-learning pytorch mental-health bilstm action-units roberta multimodal-fusion depression-detection phq8 e-daic
-
Updated
May 30, 2026 - Python