Generating image captions describing images using transformer-based models

Image caption creation, which is a very challenging task in the field of natural language processing, is called the automatic production of subtitles by the computer in the most appropriate way to the human language structure. The most important goal when creating a caption is to create sentences that describe the image in the most accurate and best way. In this regard, the use of transformer-based language models greatly affects sentence performance. In this thesis, the comparative analysis of transformer-based language models used in generating captions from images, the performance of convolution-based language models, and deep learning methods are examined. Within the scope of this determined goal, we have four applications developed on two different data sets. In our applications, sentences in English language describing the images are produced by using MSCOCO and Flickr8k datasets.