Sign Language Detection Using YOLOv10

Communication barriers pose significant challenges for individuals who are deaf, impacting their ability to interact seamlessly in society. American Sign Language (ASL) serves as a vital communication medium within this community, yet it remains largely unfamiliar to the general populace. This gap hampers access to essential services and social interactions. Leveraging advancements in deep learning and computer vision technologies, this project introduces a real-time ASL letter detection system utilizing the YOLOv10 model, renowned for its efficiency and accuracy in object detection. The system captures live video feed via a webcam, detects ASL letters, and provides visual feedback through annotated video frames. Our model, trained on a custom dataset encompassing diverse hand positions, orientations, and lighting conditions, demonstrates high accuracy and robustness. The methodology includes data collection and preprocessing, model architecture and training, and real-time detection implementation. Evaluation metrics such as precision, recall, mean Average Precision (mAP), and accuracy affirm the model’s effectiveness, with an accuracy of 92.86%. This system offers significant potential for applications in education, healthcare, and public services, promoting inclusivity and accessibility. Future work aims to enhance the model’s performance by expanding the ASL dataset’s size and diversity, thereby improving its ability to generalize across various real-world scenarios.