2025/04/01 by Daniil Pakhomov, Xie, Liangbin, Pakhomov, Daniil +18 · 4 citations
Computer Science · #Adapter (computing) #Adversarial system #Benchmark (surveying) #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Image (mathematics) #Inpainting #Multimodal Machine Learning Applications #Scheme (mathematics)
paper · pdf · doi:10.48550/arxiv.2504.00996
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2025/04/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
This paper introduces TurboFill, a fast image inpainting model that enhances a few-step text-to-image diffusion model with an inpainting adapter for high-quality and efficient inpainting. While standard diffusion models generate high-quality results, they incur high computational costs. We overcome this by training an inpainting adapter on a few-step distilled text-to-image model, DMD2, using a novel 3-step adversarial training scheme to ensure realistic, structurally consistent, and visually harmonious inpainted regions. To evaluate TurboFill, we propose two benchmarks: DilationBench, which tests performance across mask sizes, and HumanBench, based on human feedback for complex prompts. Experiments show that TurboFill outperforms both multi-step BrushNet and few-step inpainting methods, setting a new benchmark for high-performance inpainting tasks. Our project page: https://liangbinxie.github.io/projects/TurboFill/