Diffusion-based Image Stylization

Student: Yan Maksimov

Educational Programme: Applied Mathematics and Information Science (Bachelor)

Final Grade: 9

Year of Graduation: 2024

Nowadays, large text-to-image models have rapidly gained prominence across creative fields, generating visually compelling outputs from textual prompts. However, there is also the necessity of editing these images. With different styles and other required semantics, existing methods require full parameter fine-tuning or fine-tuning a small percentage of them. However, even PEFT setup requires relatively huge computational costs for these tasks, apart from the other bunch of problems, among which there is language drift and severe knowledge degradation. One of the most popular approaches in this area is to utilise the in-context abilities of large diffusion models. In this work, we propose new in-context methods called Soft Style Aligned and Naive Soft Style Aligned that outperform direct ancestor Style Aligned.

Full text (added May 20, 2024)

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses