Automatic Face Image Labelling of Ethnicity, Emotion, Age, and Gender

Research Focus

Summary

To build accurate computer vision models, large and trustworthy datasets are necessary. This paper suggests an Automatic Facial Image Labeling System built upon semi-supervised learning methods using pre-trained models. The framework enables automatic and correct labeling of facial pictures, significantly increasing the speed of generating datasets, enhancing consistency, and minimizing human bias. The proposed model performs well in age and gender classification and demonstrates the potential of semi-supervised learning to reduce manual labeling efforts for multi-attribute facial recognition.