/Diffusion-Project

In this project, we propose a system to generate illustrations for paragraphs from the books in a specific style based on their genres. The pipeline of the system consists of 3 main parts: knowledge distillation-based information retrieval model to summarize a given paragraph, a cluster model to figure out the genre, and fine-tuned stable diffusion

Primary LanguageJupyter Notebook

Diffusion-Project

The visuals in our childhood books are some of the most memorable components, so why do we not see them more often in adult novels as well? One may explain it as lack of style or devotion to the force of the word, or simply that advanced readers do not require illustrations. However, it is diffi- cult to argue that an adult text with some artwork supporting the story complements a book. There are many reasons why we should use more images in books. Images can provide additional informa- tion that supplements the text. They are helpful for communicating ideas and information, and by supporting the text’s material, they can enhance understanding. They can break up large blocks of text, making the book more visually appealing. They can help set the tone and atmosphere of the book so the reader doesn’t lose attention. All these reasons compelled us to make a solution to make visualization of text paragraphs in novels to make them even more interesting

I deleted initial dataset 'books_and_genres.csv' since it weights 3GB. If you need it contact me amirbek.djanibekov@gmail.com