DeepStableYolo: deepseek-driven prompt engineering and search-based optimization for AI image generation

• Autores: Hector D. Menéndez, Gema Bello Orgaz, Cristian Ramírez Atencia
• Localización: Actas del XVI Congreso Español de Metaheurísticas, Algoritmos Evolutivos y Bioinspirados: (MAEB 2025) 28-30 de mayo, Donostia/San Sebastián / coord. por Leticia Hernando Rodríguez, Josu Ceberio Uribe, Jon Vadillo Jueguen, 2025, ISBN 978-84-1319-656-5, págs. 61-70
• Idioma: inglés
• Enlaces
  • Texto Completo Libro
• Resumen

  • I-driven image generation heavily relies on effective prompt engineering and precise tuning of model parameters. The StableYolo framework addressed these challenges by integrating evolutionary computation with Stable Diffusion, enabling simultaneous optimization of both prompts and model parameters while using YOLO as a guiding metric to enhance image quality. In this work, we extend the capabilities of StableYolo by introducing mechanisms for prompt improvement through large language models (LLMs), aiming to maximize image generation quality. We incorporate DeepSeek to enhance prompt engineering, ensuring more effective and context-aware prompt generation. However, our refined approach demonstrates that enhancing prompts does not yield significant improvements in either the efficiency or quality of AI-generated images, suggesting that clear and concise prompts are equally effective in the process.