A Preliminary Study on the Possibility of Scene Captioning Model Integration as an Improvement in Assisted Navigation for Visually Impaired Users

This research introduces a new approach to augment image captioning for visually impaired individuals by integrating depth data with RGB images. An overview of existing assistive tools and technologies indicates their limited adoption due to high costs and various constraints. The proposed model, designed to tackle these challenges, has the potential to enhance scene comprehension and navigation for the visually impaired. The model, which includes stages from data collection to its integration with assistive tools, uses a unique neural network architecture to process both image types, merge their outputs, and generate more detailed and practical descriptions of the environment. However, specific challenges exist, such as securing an appropriate RGB-D image dataset and creating an efficient neural network. Ongoing research efforts are vital to refine the model and evaluate its real-world applicability.