FoA-SR: Faithful or Aesthetic? Profile-Aware Preference Optimization for Real-World Image Super-Resolution
Researchers have developed a new approach called FoA-SR for image super-resolution that can generate distinct restoration profiles. This method allows for either faithful reconstructions that prioritize structural integrity and reference consistency, or aesthetic reconstructions that focus on visually pleasing details. The system uses a supervised SR adapter trained with various losses, then fine-tunes separate LoRA adapters using profile-specific rewards to achieve these different objectives. AI
IMPACT Enables more nuanced control over image generation, allowing users to prioritize either accuracy or visual appeal.