This research reveals improved age, gender, and race classification in facial images, suggesting significant advancements in demographic estimation technologies.
Key Points
High predictive accuracy was achieved with an MAE of 2.95 for age estimation, improving demographic prediction reliability.
Utilizing real-world datasets like UTKFace and synthetic data enabled robust performance across gender and race classification with accuracies of 98.3% and 93.1%, respectively.
The evaluation included mean absolute error and confusion matrices, emphasizing comprehensive insights into performance among diverse demographic groups.
Future enhancements could involve attention mechanisms and fairness-aware learning to address demographic bias in predictions.