Intelligence Embedded Image Caption Generator using LSTM based RNN Model

Abstract

Humans tend to extract information from everything they see be it living or non-living. This whole phenomenon motivated us to move in this direction and explore the field of computer vision and how this can be used with recurrent neural networks to generate captions from any image. By witnessing the recent increase in natural language processing-based applications; various other researchers have also worked on this concept and produced commendable results. Describing an image is not an easy task to implement, the structure and semantics of a sentence hold an important weight age in sentence formation. This paper approaches the problem of caption generation with an LSTM (Long-Short Term Memory) based RNN model and builds architecture based on the same to generate efficient and meaningful captions by training the dataset effectively. Flicker8k dataset is used to train our model and worked well. The accuracy of the model is evaluated based on standard evaluation metrics.