Deep Learning Empowered Hand Gesture Recognition: using YOLO Techniques

Hand gesture recognition has emerged as a pivotal research domain, particularly within human-computer interaction and assistive technology. This paper presents a comprehensive overview of our study, focusing on the role of YOLO (You Only Look Once) in revolutionizing real-time hand gesture recognition. We delve into the significance of data preprocessing, demonstrating its critical influence on model training and testing. Leveraging a YOLO configuration, we assembled a diverse dataset of over 5000 images, meticulously labeled using the Roboflow platform. This dataset was strategically partitioned into training, validation, and testing subsets, setting the stage for robust model evaluation.YOLO’s object detection prowess, characterized by remarkable accuracy and speed, has propelled it to the forefront of deep learning models. Dividing input images into a grid of cells, YOLO predicts bounding boxes and class probabilities, enabling the concurrent detection of multiple objects in a single pass. This unique mechanism makes YOLO ideally suited for real-time applications, exemplified by its application in hand gesture recognition.Our research showcases the adaptability and customization potential of YOLO, enabling fine-tuning for specific tasks and datasets. By harnessing YOLOv5, YOLOv6, and YOLO v8 models, we conduct an in-depth performance comparison, culminating in the endorsement of YOLOv8l for its superior accuracy and speed. As ongoing research and development continue, YOLO-based models are poised to further refine accuracy and efficiency, solidifying their role as catalysts for improved human-computer interaction and assistive technology.