YOLO-LRHG: Long Range Hand Gesture detection using YOLO with attention mechanism

The interpretation of hand gestures is crucial for effective non-verbal communication, particularly in HumanRobot Interaction (HRI). Despite the importance, existing research has mainly concentrated on short-range gesture recognition, inadequate for safe and effective interactions with mobile robots such as UAVs over longer distances. This study addresses this gap by exploring long-range hand gesture recognition using RGB images captured by webcams. We introduce the LRHG-10m dataset, which extends the original LRHG dataset with images captured at 8 to 10 meters in various indoor and outdoor environments. We evaluated seven attention mechanisms and the Ghost module for their performance in detecting small objects under low-parameter conditions. Our proposed YOLO-LRHG model integrates these mechanisms with an anchor-free approach, achieving an impressive mAP@0.5 of 98.5% with 3M parameters and 97.5% with 1.75M parameters, balancing computational efficiency, detection distance, and precision.