Texture-specific image editing with TALC-Net

This paper introduces the Texture-Aware Layout-Conscious Network (TALC-Net), a novel framework addressing the challenge of applying user-defined textures to specific image regions while preserving global spatial coherence. The architecture of TALC-Net consists of a texture encoder for capturing detailed texture features, a region localization layer for isolating masked areas, and a diffusion-based generative module to synthesize high-fidelity outputs. Additionally, an information controller ensures smooth integration of texture and semantic features, enabling context-aware and visually coherent regeneration. Extensive experiments on a curated interior design dataset with over 10,000 semantically annotated images demonstrate that TALC-Net outperforms state-of-the-art methods. It excels in maintaining layout integrity, ensuring texture fidelity, and enhancing creative flexibility, making it a powerful tool for layout-sensitive image generation tasks.