Hand Gesture Recognition System Using Transfer Learning

The field of computer vision and gesture recognition has witnessed the emergence of strong real-time detection systems, spurred by the breakthroughs in deep learning. Nonverbal cues, especially hand gestures, are extremely important in human-vehicle interaction. A novel method for recognizing and classifying hand movements associated with requesting a lift or signaling to stop a vehicle is provided in this study. By utilizing the YOLOv8 (You Only Look Once version 8) paradigm, this research develops a methodical approach to obtain dataset that include a wide variety of typical hand gesture samples in traffic situations. The hand gesture dataset created is then used to train and optimize the YOLOv8 model, allowing for accurate distinction between these two crucial hand movements. The trained model with precision, recall, and a map50 of 78.7%, 75.3%, and 73.1% potentially helps in enabling the monitoring of the driver's conduct for adherence and integrity, enhancing interactions between autonomous cars, promoting road safety, and increasing accessibility for public transportation.