Enhancing the ability to recognize facial expressions in young children: A method using few-shot learning and cross-dataset validation

Facial Emotion Recognition (FER) is an important field inc omputer vision that h as significant im plications for human-computer interaction, healthcare, and education. This study aims to tackle the particular difficulties a ssociated with identifying emotions in young children. It primarily concentrates on recording the facial expressions of infants, toddlers, and young children between the ages of 0 and 5 years old. This dataset has been personally collected. Within this dataset, the few-shot learning technique is employed to enhance the accuracy of facial expression recognition (FER), specifically addressing the unique challenges associated with early childhood expressions. The study seeks to explore the potential of few-shot learning in enhancing facial expression recognition (FER) for young children. It contributes to the broader area by conducting a literature review, evaluating methodologies, and developing a unique framework. This proposed technique achieved an outstanding accuracy of 87.49 % when applied to the obtained dataset. In addition, the validation on external datasets, such as The Japanese Female Facial Expression (JAFFE) and CK+, consistently showed excellent performance, achieving accuracies of 98.65% and 99.35% correspondingly. These data highlight the efficacy o f t he proposed technology i n various facial expression recognition challenges. This research enhances the development of sensitive facial expression recognition (FER) algorithms by highlighting subtle emotional nuances in early childhood. It also contributes to a deeper knowledge of how emotions are expressed and opens up possibilities for transformational applications in fields such as education and healthcare.