Loading...
Publication
Geração inteligente de interfaces gráficas para a Web
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 7.24 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A geração automática de interfaces gráficas para aplicações web tem vindo a afirmarse
como uma abordagem inovadora para aumentar a eficiência, a uniformidade e a qualidade
no desenvolvimento de software. Este trabalho apresenta o desenvolvimento, implementação
e validação de um sistema inteligente capaz de gerar código de frontend de forma automatizada
a partir de mockups visuais, integrando modelos de machine learning e pipelines multimodais.
O processo de investigação partiu de uma revisão sistemática da literatura, com recurso
à metodologia PRISMA, e da análise comparativa de abordagens existentes. A solução
desenvolvida assenta num pipeline multimodal que combina um modelo de visão
computacional, para interpretar os mockups, com um modelo de linguagem de grande escala
Large Language Model (LLM) adaptado através de fine-tuning para gerar o código final.
A avaliação recorreu a um pipeline de testes automatizado e a métricas objetivas de
qualidade, desempenho e produtividade, permitindo comprovar a viabilidade da automação
inteligente na geração de interfaces. Os resultados obtidos evidenciam o potencial para reduzir
o esforço manual, aumentar a consistência visual e acelerar o ciclo de desenvolvimento. Assim,
a investigação realizada estabelece uma base sólida para futuras evoluções e para a
industrialização desta abordagem em contextos organizacionais.
The automatic generation of graphical interfaces for web applications has emerged as an innovative approach to increasing efficiency, uniformity, and quality in software development. This work presents the development, implementation, and validation of an intelligent system capable of automatically generating frontend code from visual mockups, integrating machine learning models and multimodal pipelines. The research process began with a systematic literature review, following the PRISMA methodology, and a comparative analysis of existing approaches. The developed solution is based on a multimodal pipeline that combines a computer vision model to interpret mockups with a LLM adapted via fine-tuning to generate the final code. The evaluation employed an automated testing pipeline and objective metrics for quality, performance, and productivity, demonstrating the feasibility of intelligent automation in interface generation. The results highlight the potential to reduce manual effort, increase visual consistency, and accelerate the development lifecycle. Thus, this research establishes a solid foundation for future advancements and the industrialization of this approach in organizational contexts.
The automatic generation of graphical interfaces for web applications has emerged as an innovative approach to increasing efficiency, uniformity, and quality in software development. This work presents the development, implementation, and validation of an intelligent system capable of automatically generating frontend code from visual mockups, integrating machine learning models and multimodal pipelines. The research process began with a systematic literature review, following the PRISMA methodology, and a comparative analysis of existing approaches. The developed solution is based on a multimodal pipeline that combines a computer vision model to interpret mockups with a LLM adapted via fine-tuning to generate the final code. The evaluation employed an automated testing pipeline and objective metrics for quality, performance, and productivity, demonstrating the feasibility of intelligent automation in interface generation. The results highlight the potential to reduce manual effort, increase visual consistency, and accelerate the development lifecycle. Thus, this research establishes a solid foundation for future advancements and the industrialization of this approach in organizational contexts.
Description
Keywords
Automatic Interface Generation Machine Learning Angular Frontend Automation Mockups Geração automática de interfaces