Devdiscourse

The AI pipeline extracted valence (positive or negative emotion) and arousal (emotional intensity) features from participants’ facial expressions using two convolutional neural networks trained on the large-scale AffectNet dataset and fine-tuned for the task. These emotion-derived features were then analyzed using three machine learning models, K-Nearest Neighbors, Logistic Regression, and Support Vector Machine, to classify participants by cognitive status. A nested cross-validation framework ensured that performance metrics were unbiased despite the relatively small dataset.
See full story at Devdiscourse
Sign up for our newsletter!
Get the latest information and inspirational stories for caregivers, delivered directly to your inbox.