Enhancing Low-Resolution Images for FaceNet Facial Recognition Using GFPGAN on Raspberry Pi 5

Throughout the years, significant advancements have been made in the field of photography and facial recognition technologies. And these technological advancements have become essential for numerous applications. However, traditional image enhancement algorithms often struggle to improve the image quality required for reliable feature extraction in facial recognition. In response, this study proposes the use of a Generative Adversarial Network (GAN), specifically GFPGAN (Generative Facial-Prior GAN), to enhance low-resolution images for facial recognition using FaceNET on Raspberry Pi 5. The system utilizes a Raspberry Pi V3 Camera Module to capture the images of participants at varying distances and angles. A total of 500 images underwent recognition, 250 of which were enhanced by GFPGAN. The system achieved an overall accuracy of 86.80% with GFPGAN-Assisted Facial Recognition, outperforming Standard Facial Recognition with an accuracy of 57.60 %. The results of the study emphasize the capability of GFPGAN in enhancing low-resolution images for facial recognition, which enables the development of facial recognition systems that can maintain high accuracy under suboptimal imaging conditions and the development of more dependable and secure applications that concern privacy and security.