Deep Learning Framework for Sign Language Recognition Using Inception V3 with Transfer Learning

Sign language recognition poses unique difficulties due to the complexities of hand gestures and movements. Real-world applications often require robust models capable of handling diverse environmental conditions and variations in signing styles. Transfer learning, a technique that utilizes knowledge from pre-trained models, has emerged as a promising method to address these challenges. Although transfer learning has been successful in various visual recognition tasks, its application to American Sign Language (ASL) recognition remains relatively unexplored. This research showcases the effectiveness of combining transfer learning with advanced CNN architectures like Inception V3 for American Sign Language (ASL) recognition. Through fine-tuning Inception V3, adjusting learning rates, applying data augmentation, and rigorous evaluation, the model achieved an impressive 99.12% accuracy on test datasets. These findings represent a significant improvement in ASL recognition accuracy compared to baseline models. The success of this method underscores the importance of fine-tuning and data augmentation in advancing accessible and reliable sign language recognition technology for practical use.