Author : Ms. Priyanka Raut and Mrs. Rushali A Deshmukh

Automatic Image Captioning has gained significant interest and aims towards generating Natural language sentences based on the given input image which is a very challenging task. It has been imagined that machines one day will understand the visual world at a human degree of knowledge using Artificial Intelligence. Although there are various solutions been generated by the Neural Networks in recent times which have given precise results, still a major problem arises is that the algorithm can only describe the concepts seen in the training data set with more error rate captions. In this paper, the proposed methodology consists of the Deep Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) algorithm for generating natural language captions based on Images with more accurate results. We evaluate our methodology on Flickr8k data set.

Keywords: Image Pre-processing, CNN Algorithm, Feature Extractors, Feature Vector, RNN, LSTM, Caption Generation