Optimised YOLO11 for real-time sign language detection in 6 G-enabled metaverse environments for inclusive enterprise communication and training

This is the authors' abstract. We don't add key points for this paper.

The integration of 6 G communication and artificial intelligence transforms real-time interactions in immersive environments, such as the metaverse. This paper proposes an optimised YOLO11 real-time sign language detection model to support inclusive communication and training in enterprise metaverse applications. The model’s performance is enhanced through hyperparameter tuning using a genetic algorithm, optimising factors like learning rate, momentum, and data augmentation. The model achieves a 0.925 precision, 0.865 recall, 0.9368 mAP50, and 0.8065 mAP50–95, outperforming the baseline YOLO11 and earlier YOLO versions. Lower GFLOPS and reduced parameters (2.59 M) enable deployment on lightweight metaverse and enterprise devices.