Image Caption Generation Using Deep Learning

Abstract

From the perspective of humans and computers, a picture can be interpreted in distinct manner. In the case of humans, a picture will be clearly a few description or scene of a motion or environment and so forth, whilst with respect to computers, it is just a few aggregates of pixels or digital numbers. The system of photo captioning offers with assigning inner facts in the shape of captions with the aid of extracting the applicable functions from an input picture. This venture aims at producing meaningful captions for a given picture. The proposed work is based on deep neural networks. The proposed work has three fundamental units. The first is the picture module that is given as input to the function extractor unit. The next unit is a feature extractor unit based on CNN (Convolutional Neural Network) which extracts the applicable characteristic. The final unit is the language generator. It generates sentences that describe the input image. To assess the quality of the generated textual content, BLEU(Bi-Lingual Evaluation Understudy) rating is used. Suitable captions will help the users to search snapshots with lengthy queries. Such systems may also be beneficial for visually impaired humans in understanding pictures.