Research map / Image generation / Editing and style transfer
MaskSketch: Unpaired Structure-Guided Masked Image Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use attention-based structural guidance with a pretrained MaskGIT generator instead of unpaired translation baselines | MaskSketch · ImageNet-Sketch 10-class | FID: 68.65 -> 33.24 | Helped |
| Use attention-based structural guidance with a pretrained MaskGIT generator instead of unpaired translation baselines | MaskSketch · Pseudosketches 10-class | FID: 93.23 -> 56.55 | Helped |
| Use attention-based structural guidance with a pretrained MaskGIT generator instead of the strongest listed baseline | MaskSketch · ImageNet-Sketch 10-class | User preference: 19.78% -> 59.35% | Helped |
| Use attention-based structural guidance with a pretrained MaskGIT generator instead of the strongest listed baseline | MaskSketch · Pseudosketches 10-class | CLIP prompt: 45.67 -> 59.48 | Helped |
| Add CLIP-based rejection sampling, selecting one output from multiple trials (ablation) | MaskSketch · ImageNet-Sketch 10-class | FID: 34.23 -> 33.24 | Helped |
| Add CLIP-based rejection sampling, selecting one output from multiple trials (ablation) | MaskSketch · ImageNet-Sketch 10-class | CLIP prompt: 71.87 -> 67.10 | Hurt |
| Add CLIP-based rejection sampling, selecting one output from multiple trials (ablation) | MaskSketch · Pseudosketches 10-class | FID: 60.44 -> 56.55 | Helped |
| Add CLIP-based rejection sampling, selecting one output from multiple trials (ablation) | MaskSketch · Pseudosketches 10-class | CLIP feature distance: 26.85 -> 25.60 | Helped |
| Use early transformer attention layers for structural guidance rather than final layers (ablation) | MaskSketch · Image generation from sketches | Structural fidelity | Mixed |
| Use final transformer attention layers for structural guidance rather than early layers (ablation) | MaskSketch · Image generation from sketches | Appearance diversity | Helped |
| Stop sampling at mask rate 0.25 rather than continuing the schedule (ablation) | MaskSketch · Sketch-to-photo translation | Realism | Helped |
Closest papers by meaning
- A Simple Approach to Unifying Diffusion-based Conditional Generation (ICLR 2025 · similarity 0.87)
- Halton Scheduler for Masked Generative Image Transformer (ICLR 2025 · similarity 0.87)
- Mask^2DiT: Dual Mask-based Diffusion Transformer for Multi-Scene Long Video Generation (CVPR 2025 · similarity 0.87)
- SmartMask: Context Aware High-Fidelity Mask Generation for Fine-grained Object Insertion and Layout Control (CVPR 2024 · similarity 0.86)
- Zero-Painter: Training-Free Layout Control for Text-to-Image Synthesis (CVPR 2024 · similarity 0.86)
- Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis (ICLR 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.