Machine Learning Based Sign Language Recognition Translation Methods

In many computer vision applications, object recognition is essential, and the YOLO (You Only Look Once) series has led the way in developing real-time object detection models. YOLOv5, the most recent version of the YOLO architecture, is used in this study to tackle the particular problem of Indian Sign Language (ISL) identification. The dataset, which can be exported via roboflow.com, includes a variety of Indian Sign Language gesture examples. The strong architecture of YOLOv5, which consists of a strong neck, head, and backbone, is used for precise and effective detection. Taking into account the unique characteristics of ISL gestures and the need for instantaneous inference, using the dataset to train the model. Our study explores the components of YOLOv5 and its modifications for the ISL environment, delving into the complexities of its design. The training procedure involves a meticulous evaluation of hyperparameters, dataset attributes, and sign language detection-specific problems. The model’s ability to precisely identify and locate ISL gestures in practical settings is shown by the results. YOLOv5’s strengths are revealed via comparison evaluations with other object identification algorithms, and metrics including accuracy, precision, and recall are presented.