Comparative Analysis of YOLOv5 and YOLOv8 for Real-Time Hand Gesture Recognition

Hand gestures recognition technology enhances human-computer interaction in applications such as sign language interpretation and contactless interfaces. This research focuses on the comparative analysis of YOLOv5 and YOLOv8 models for recognizing hand gestures of digits 1 to 5. Both models underwent training on a custom dataset of 9,680 labeled images without any additional differing variables. YOLOv8's performance was better, achieving a mAP@0.5 of 0.995, precision 0.996, recall 0.992, and F1 score 0.994, compared to YOLOv5's 0.981, 0.983, 0.975, and 0.979. In addition, YOLOv8's faster inference of 7.1 ms vs 8.3 ms and smaller model size of 23 MB make it more suited for real-time usage. Further analysis through confusion matrix showed stronger endorsement of YOLOv8's claim of better class separation and discrimination even in the most difficult scenarios. These findings confirm YOLOv8's superiority as the more dependable model for real-time hand gesture recognition with precision and speed.