Indian Sign Language Classification Using Transfer Learning

In recent years, there has been increased interest in the field of machine learning to develop accurate and efficient Indian Sign Language (ISL) classification models. This interest stems from the urgent need to overcome communication barriers and promote the inclusion of the hearing impaired. Here, ISL classification using transfer learning techniques with both a conventional convolutional neural network (CNN) and the powerful ResNet50 architecture has been used. Transfer learning has become a promising strategy in machine learning to enable models to be ready trained to adapt large datasets with limited labelled data to new tasks. The methodology involves fine-tuning models pre-trained on the ISL dataset using knowledge learned from different visual concepts. By exploiting the learned features of these deep learning architectures, this methodology can effectively capture the complex patterns and nuances of ISL gestures. Experimental results demonstrate the effectiveness of transfer learning in ISL classification tasks. Both custom CNN and ResNet50 architectures achieve significant accuracy, which demonstrates the robustness and generalization of the transmitted data. These high accuracy rates highlight the potential of transfer learning as a valuable tool to solve the unique challenges of ISL classification. In addition, it is presented here as a user-friendly web-based application built with Stream lit to facilitate the smooth use of trained models. This application provides a practical platform for users to interact with the ISL classification system, enabling real-world applications and promoting accessibility for the hard of hearing.