Open-Set Profile-to-Frontal Face Recognition on a Very Limited Dataset

Open-set face recognition on a small dataset with limited image samples per individual poses significant challenge and is topic of active research. Therefore, this study investigated the problems open-set verification identification known as ITB Frontal Profile Limited Dataset (IFPLD), which included only one frontal profile individual. Various training procedures were used to obtain more appropriate network embedding for feature representation dataset. Transfer learning was employed improve performance models by fine-tuning networks using properties similar those IFPLD. The results showed that SimCLR method generated optimal Siamese network. prototypical an N-way-k-shot scenario where k-1 came from data augmentation outperformed maximum 17.0% accuracy improvement. transformation 1-shot k-shot critical achieving high performance.