ARCHIVES

The visually impaired individuals are forced to struggle to read pictorial information such as pictures, objects and the surrounding environment and hence limit the degree of self-reliance and access to internet information. Within the framework of this paper, the authors propose to use an AI-based image description system that may generate meaningful textual descriptions of pictures automatically and helps blind people to have a better perception of visual worlds and online media. Convolutional Neural Networks (CNNs) are applied to extract visual features and process the discovered visuals through sequence-based NLP models to produce captions that can be understood by human beings. The output of the descriptive generated texts is then converted into a speech by the text-to-speech software that provides real-time and hands-free assistance by way of a web or mobile interface. This system aims at improving the quality of life, situational awareness and accessibility of the visually impaired persons.