Indian Sign Language classification and Image Augmentation using GAN

A significant portion of the deaf and hard of hearing population in India communicates primarily through Indian Sign Language (ISL). Precise and successful recognition of ISL (International Sign Language) gestures can dramatically improve communication between people with hearing impairments and the rest of the community. This research introduces a novel method for identifying ISL motions that combines Convolutional Neural Networks, or CNN, with Generative Adversarial Networks (GAN). The study presents a unique method for classifying static hand positions representing letters in sign language using RGB photographs that employ a densely linked Convolutional Neural Network (CNN) design. The model employed a generative adversarial network (GAN) to balance the dataset to generate synthetic data that looked like actual data. With the help of GAN, the model has also augmented data since it was limited. Our proposed model demonstrates promising results in accurately detecting and classifying ISL gestures with an accuracy of 97.69 %, contributing to improved accessibility and inclusivity.