BordadoBOT
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
IPCB. EST
Abstract
O setor das artes e do design, especialmente no contexto do bordado, possui uma rica tradição que reflete estilos e técnicas específicas, como o bordado de Castelo Branco. Com o avanço da tecnologia e da inteligência artificial, surgem novas formas de expressão artística, permitindo a reinterpretação inovadora de estilos tradicionais. Este projeto explora a adaptação e criação de imagens ao estilo do bordado de Castelo Branco utilizando técnicas de inteligência artificial, investigando a interseção entre arte tradicional e tecnologia moderna. Para isso, foram aplicados modelos de Stable Diffusion, capazes de gerar imagens que reflitam as
características visuais deste bordado. O trabalho incluiu uma revisão do estado da arte sobre transferência de estilos artísticos com inteligência artificial, a criação de um dataset de imagens do bordado de Castelo Branco e o treino de diversos modelos. Durante o desenvolvimento, foram realizados testes exaustivos com diferentes parâmetros de geração e otimização dos prompts textuais, com o objetivo de maximizar a qualidade e fidelidade do estilo nos resultados. Essa abordagem permitiu identificar as configurações que mais se adequavam para criar imagens que remetem ao estilo do bordado, tanto no modo txt2img quanto img2img. Por fim, foi desenvolvido um protótipo multiplataforma, acessível e intuitivo, que permite a utilizadores com diferentes níveis de conhecimento técnico interagir com os modelos treinados e gerar imagens no estilo pretendido. O projeto não apenas preserva e celebra a herança cultural do bordado de Castelo Branco, mas também inova ao integrar inteligência artificial, possibilitando a aplicação desse estilo tradicional a novas imagens e contextos, além de contribuir para a facilitação do acesso a ferramentas criativas baseadas em inteligência artificial
Abstract: The arts and design sector, particularly in the context of embroidery, has a rich tradition that reflects specific styles and techniques such as Castelo Branco embroidery. With the advancement of technology and artificial intelligence, new forms of artistic expression have emerged, enabling the innovative reinterpretation of traditional styles. This project explores the adaptation and creation of images in the style of Castelo Branco embroidery using artificial intelligence techniques, investigating the intersection between traditional art and modern technology. To achieve this, Stable Diffusion models were applied to generate images that reflect the visual characteristics of this embroidery. The work included a review of the state of the art in artistic style transfer with AI, the creation of a dataset of Castelo Branco embroidery images, and the training of several models. During development, extensive tests were conducted with different generation parameters and prompt optimization, aiming to maximize the quality and stylistic fidelity of the results. This approach allowed the identification of the most suitable configurations for generating images that evoke the embroidery’s style, both in txt2img and img2img modes. Finally, a multiplatform, accessible, and intuitive prototype was developed, enabling users with different technical backgrounds to interact with the trained models and generate stylized images. The project not only preserves and celebrates the cultural heritage of Castelo Branco embroidery, but also innovates by integrating artificial intelligence, making it possible to apply this traditional style to new images and contexts, while contributing to the democratization of access to creative AI-based tools.
Abstract: The arts and design sector, particularly in the context of embroidery, has a rich tradition that reflects specific styles and techniques such as Castelo Branco embroidery. With the advancement of technology and artificial intelligence, new forms of artistic expression have emerged, enabling the innovative reinterpretation of traditional styles. This project explores the adaptation and creation of images in the style of Castelo Branco embroidery using artificial intelligence techniques, investigating the intersection between traditional art and modern technology. To achieve this, Stable Diffusion models were applied to generate images that reflect the visual characteristics of this embroidery. The work included a review of the state of the art in artistic style transfer with AI, the creation of a dataset of Castelo Branco embroidery images, and the training of several models. During development, extensive tests were conducted with different generation parameters and prompt optimization, aiming to maximize the quality and stylistic fidelity of the results. This approach allowed the identification of the most suitable configurations for generating images that evoke the embroidery’s style, both in txt2img and img2img modes. Finally, a multiplatform, accessible, and intuitive prototype was developed, enabling users with different technical backgrounds to interact with the trained models and generate stylized images. The project not only preserves and celebrates the cultural heritage of Castelo Branco embroidery, but also innovates by integrating artificial intelligence, making it possible to apply this traditional style to new images and contexts, while contributing to the democratization of access to creative AI-based tools.
Description
Keywords
Inteligência artificial, Deep learning, Transferência de estilo, Modelos de difusão, Arte generativa, Artificial intelligence, Style transfer, Diffusion models, Generative art
Citation
PIRES, João Rafael de Matos ; CHARRO, Gabriel Horta (2025) - BordadoBOT. Castelo Branco: IPCB. EST. 181 p. Relatório do Trabalho de Fim de Curso de Engenharia Informática.