Image Caption Generation entails the creation of natural language descriptions for images. Due to the complexity of the visual content as well as the semantic details of the corresponding natural language, a problem arises. In this project, we propose a system for generating image captions by integrating the architecture of Convolutional-Neural-Networks along with Long-Short-Term Memory networks. The methodology involves the CNN gathering of visual features followed by the LSTM generation of captions. We also incorporate attention mechanisms to enhance the performance of the model by enabling it to concentrate on pertinent visual features while generating captions. Using CIDEr, METEOR and BLEU scores, for evaluation of the performance of our modified and restructured model using standard benchmark datasets (Flickr8k) and BLEU, CIDEr, and METEOR scores. Our proposed system has the potential for use in the image retrieval, the image description, and other multimedia applications requiring image analysis and natural language processing.

Convolutional-Neural-Network, BLEU, CIDEr and METEOR score, Long-Short-Term-Memory, Image Captioning.