GMC-YOLO: A Face Detection Algorithm Based on Enhanced YOLOv8

Face detection represents a popular application in the field of artificial intelligence. This paper addresses challenges specific to scenarios where cameras are positioned at a considerable distance and angle, resulting in issues such as small or obstructed faces, as well as detection failures or errors caused by small-scale faces. In response, we propose the GMC-YOLO face detection algorithm based on YOLOv8. The introduction of the GMConv module in the feature fusion section allows for enhanced extraction of facial features across different scales. Additionally, to address indoor face overlap issues, the Soft-NMS algorithm is incorporated into our model. Face detection experiments conducted on the Wider-Face dataset demonstrate that, compared to alternative algorithms, GMC-YOLO exhibits superior accuracy and commendable real-time performance.