Deep Learning based Audio Description of Visual Content by Enhancing Accessibility for the Visually Impaired

Individuals with vision impairments face significant barriers in accessing and understanding visual content, hindering their full engagement in daily activities. However, technological advancements offer solutions to enhance accessibility for the blind and visually impaired. Text-to-speech technology converts written text into spoken words, enabling visually challenged individuals to listen to books and articles. Moreover, image recognition technology, like the CNN-RNN model, generates audio descriptions of photographs, aiding in visual comprehension. As technology evolves, more innovative solutions are being developed to support inclusivity for the visually impaired. In order to solve this problem, images are converted into textual descriptions, and then a text-to-speech API is used to turn them into audio in the proposed work. By leveraging the power of technology and achieving positive results with the Flickr8k dataset, a more inclusive world is contributed to by us. This technique enables visual content that would otherwise be challenging for them to be accessed and understood by visually challenged individuals. The recognition of objects, people, and context in images is proven effective in generating image captions by utilizing the CNN-RNN model. The performance of the suggested model is evaluated using the BLEU score and compared to existing models. While relatively low scores are exhibited by VGG16 and RESNET, BLEU scores ranging from 70% to 100% are achieved using the proposed Inception V3model.