Image Recognition and Voice Translation for Visually Impaired

Abstract

Image recognition and voice translation for the visually impaired is principally centered on aiding the visually impaired individuals which can guide them through changing the image into voice. Image captioning is that generating the caption of the given image. The arrival of image recognition and voice translation system can facilitate visually impaired individuals to visualize the globe. Image recognition needs to understand the numerous things, their properties, and their connections during an image. The “Flickr_8k” dataset is employed for this project with having 8000 pictures. During this project, we have a tendency to principally specialize in the eye mechanism, that extracts the options through VGG16 and converts the options into language through LSTM. During this project, we've projected a changed LSTM which can enhance the addition of data that have a tendency to get from a picture as further input to each unit of the block, which enhances the higher results on the recognition and translation. The sentence that's generated from the image is evaluated exploitation blue cheese Score. Moreover, the benefits and therefore the short comings of the prevailing ways and projected technique area unit mentioned and therefore the analysis criteria has been non commissioned. Hence, we have a tendency to area unit developing this model that generates a caption for the given image and that we show our model advances the state of art on tasks that need the joint process of Image and sentence description and finally converting it into voice.