Research map / Image generation / Text-to-image diffusion
High-Resolution Artwork Outpainting with Global Blueprint Guidance and Layout Control
Junha Kim, Hyunjoon Park, Donghyeon Cho
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add global guidance feature injection to Stage 2 (ablation) | BlueOut · Artwork outpainting ablation | FID: 22.2319 -> 9.3064 | Improved |
| add patch position token (ablation) | BlueOut · Artwork outpainting ablation | FID: 17.9578 -> 9.3064 | Improved |
| add forward-diffusion blueprint initialization (ablation) | BlueOut · Artwork outpainting ablation | FID: 9.6048 -> 9.3064 | Improved |
| use blueprint-guided Stage 2 instead of latent blending only (ablation) | SD Inpainting (parallel) · Artwork outpainting ablation | FID: 37.2675 -> 9.3064 | Improved |
| add layout conditioning (ablation) | BlueOut Stage 2 · Artwork outpainting layout ablation | layout AP: 0.3973 -> 0.4336 | Improved |
| add layout conditioning (ablation) | BlueOut Stage 2 · Artwork outpainting layout ablation | FID: 9.7111 -> 9.3064 | Improved |
| use Stage 2 synthesis instead of super-resolving the Stage 1 blueprint (DRCT-L) (ablation) | BlueOut · Stage 1 + SR | FID: 22.0840 -> 9.3064 | Improved |
| increase guidance scale from 3.0 to 7.5 (ablation) | BlueOut · Guidance scale | FID: 9.3064 -> 12.9130 | Hurt |
| weaken forward-init strength from 1.0 to 0.50 (ablation) | BlueOut · Forward-init strength | layout AP: 0.4336 -> 0.4979 | Mixed |
| use parallel Stage 2 on 8 GPUs vs ProOut | BlueOut · Artwork outpainting (main) | inference time (s): 18.31 -> 7.50 | Improved |
| use BlueOut instead of ProOut at 600% expansion | BlueOut · Expansion ratio 600% | FID: 89.84 -> 81.86 | Improved |
| run Stage 2 on a single GPU (ablation) | BlueOut · Single-GPU inference | inference time (s): 15.73 s vs 18.31 s ProOut, slower than SD Inpainting 13.43 | Mixed |
Closest papers by meaning
- Infinite-Canvas: Higher-Resolution Video Outpainting with Extensive Content Generation (AAAI 2025 · similarity 0.88)
- TD-Paint: Faster Diffusion Inpainting Through Time-Aware Pixel Conditioning (ICLR 2025 · similarity 0.87)
- Continuous-Multiple Image Outpainting in One-Step via Positional Query and A Diffusion-based Approach (ICLR 2024 · similarity 0.86)
- HD-Painter: High-Resolution and Prompt-Faithful Text-Guided Image Inpainting with Diffusion Models (ICLR 2025 · similarity 0.86)
- RoomPainter: View-Integrated Diffusion for Consistent Indoor Scene Texturing (CVPR 2025 · similarity 0.86)
- DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
