Keyword driven image description generation system

Image description generation is an important area in Computer Vision and Natural Language Processing. This paper introduces a novel architecture for an image description generation system using keywords. The proposed architecture uses a high-level feature such as keywords for generating captions. The important component of caption generation is the deep Bidirectional LSTM network. The space and computational complexity of the system are smaller than that of the CNN feature-based image description generation system. The number of parameters is also small in the keyword-based image description generation system. It generates novel meaningful sentences for images. The systems performance depends on the keyword extraction system.