Enhancing Accessibility: Image Captioning for Visually Impaired Individuals in the Realm of ECE Advancements

In the world we are living in the most important means of taking information and making sense from it is vision. According to the research made by the scientist's vision is described as our dominant sense. It is challenging for the person who doesn't have this crucial ability and such people are called blind. Blind people do lead regular lives and have unique ways of carrying out tasks, but they undoubtedly struggle, so to overcome the challenges faced by them there have been several solutions proposed and one of them is Image Captioning i.e. using deep learning models to create captions for images which shows surrounding of the blind person. The traditional architecture for generating captions is using CNN to picture captioning (YOLO - LSTM) and (CLIP - GPT). Using a common metric (BLEU), LSTM encoder-decoder model. The proposed method utilizes the novel Image Captioning approaches for Visually Impaired People data set containing several use cases (pedestrian crossing, currency detection, bus stops, stair case … etc.) to train and test through modern approach for Image captioning. The suggested method makes use of a data set with multiple use cases (pedestrian crossings, money detection, bus stops, staircases, etc.) from the innovative Image Captioning for Visually Impaired Dataset in order to train and evaluate a contemporary strategy for Image captioning. By training it on the aforementioned dataset, we examine the effectiveness of two contemporary techniques the suggested method's effectiveness is assessed.