Analyzing the Impact of Instagram Filters on Facial Expression Recognition Algorithms

The human face conveys lots of information through facial expressions to other human beings and computers, sharing information about intentions, emotional health status, age, and ethnicity. This results in the criticality of Facial Expression Recognition (FER) systems in various applications, including emotion detection, behavior analysis, human-computer interactions, etc. However, the increasing prevalence of beauty filters in social media platforms like Instagram and other platforms has raised concerns about their potential impact on the accuracy and reliability of FER systems. The beauty filters could lead to a misalignment between the facial expressions captured by FER systems and the actual emotional state of the individual. In this research, we comprehensively analyze the influence of Instagram filters on FER systems using 30 commonly used Instagram Filters and their implications for research and application on the Real-world Affective Faces Database (RAF-DB) dataset. To assess the effectiveness of the suggested approach, we performed experiments employing three distinct FER models (EfficientFace, MA-Net, and POSTER) to validate the performance. We evaluated the performance of the proposed method by employing widely recognized evaluation metrics, including accuracy, precision, recall, and F1-score. This research highlights the challenges and implications associated with Instagram filters’ influence on facial expression recognition, emphasizing the need for further research, algorithmic advancements, and ethical considerations in this evolving landscape.