Enhancement in web accessibility for visually impaired people using hybrid deep belief network -bald eagle search

In the modern era, accessing the web is the major task for Visually Impaired (VI) people that creates opportunities to connect with social media as part of professional, political, and social life. One of the difficulties faced by VI users is accessing and understanding images online. The advancements in assistive technologies based on computer-vision (CV) assist the VI people in diverse scenarios such as grocery shopping, generation of alternative text, object recognition, understanding text documents, identifying people etc. In this study, to make the digital platform user-friendly for VI people, an automated system is developed to generate alternative (alt) text for online images that are not captioned or the alt text is not specified for the image. The proposed Deep Belief Network - Bald Eagle Search (DBN-BES) method offers an effective way for VI that allows automatic captioning of web images. Our proposed work consists of two stages. The initial stage is the selection of images that are not captioned, and this selection process is obtained using the Bald Eagle Search (BES) Algorithm. After the selection stage, alt text for corresponding images is produced with the help of the Deep Belief Network (DBN) model. Thus, the proposed DBN-BES model automatically generates alt text which helps the VI people to understand the image content better. The presented model routinely increases web accessibility, addresses massive image data being created daily, and makes the web accessible to multiple users around the globe.