基於空間增強型狀態空間模型的圖像修復：SEM-Net
本文提出了一種名為 SEM-Net 的新型圖像修復模型，該模型採用空間增強型狀態空間模型（SSM）有效捕捉圖像中的長距離依賴關係和空間一致性，進而實現高品質的圖像修復。
Enhancing Image Inpainting with Vision Transformer-Based Pre-Processing for Improved Mask Representation
This research proposes a novel pre-processing methodology for image inpainting that leverages the Vision Transformer (ViT) to replace the traditional binary mask with a feature-rich representation, leading to enhanced inpainting performance across various models and datasets.
A Task-Decoupled Image Inpainting Framework for Building Class-Specific Object Removers
Decoupling object removal and restoration tasks in image inpainting networks leads to superior class-specific object removal, as demonstrated by a novel framework that leverages separate models for each task and a data curation method.
Image Inpainting via Dual Generation of Texture and Structure
The core message of this paper is that image inpainting can be effectively achieved by jointly modeling structure-constrained texture synthesis and texture-guided structure reconstruction in a two-stream network architecture, which allows the two subtasks to better leverage each other for more plausible generation.
BrushNet: A Novel Image Inpainting Model with Dual-Branch Design
BrushNet introduces a dual-branch design for image inpainting, enhancing image quality, masked region preservation, and text alignment.
BrushNet: A Plug-and-Play Image Inpainting Model with Decomposed Dual-Branch Diffusion
BrushNetは、画像修復に革新的なアプローチを導入し、優れた性能を提供します。
Towards Context-Stable and Visual-Consistent Image Inpainting: ASUKA Framework
ASUKA framework enhances image inpainting by achieving context-stability and visual-consistency through alignment with a frozen SD model.
BrushNet: A Novel Image Inpainting Model with Dual-Branch Diffusion
BrushNet introduces a dual-branch design for image inpainting, enhancing coherence and quality. The model's architecture allows for pixel-level masked image feature insertion, leading to superior performance across various metrics.
© 2024 by Linnk AI