Image captioning using xception

Abstract

The main aim of the project was to create a model that generates the captions. The captions must be syntactically correct and semantically true. Also the goal was to create a new model that can achieve outstanding results compared to other models. The data set used was Flickr8k and was trained over pre-trained Xception model. The outcomes were compared against the ResNet50 and InceptionV3 results. The evaluation metric used was BLEU (Bilingual Evaluation Understudy) Score. The BLEU Score graphs shows that Xception has outperformed the ResNet50 and InceptionV3 models.