Research map / Image generation / Editing and style transfer
Example-Guided Image Synthesis using Masked Spatial-Channel Attention and Self-Supervision
Haitian Zheng, Haofu Liao, Lele Chen, Wei Xiong, Tianlang Chen, Jiebo Luo
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 MSCA module instead of global average pooling (ablation) | Ours full · COCO-Stuff (retrieving) | FID: 101.74 -> 91.91 | Improved |
| add MSCA module instead of global average pooling (ablation) | Ours full · COCO-Stuff (retrieving) | PSNR: 15.85 -> 16.65 | Improved |
| add spatial and channel attention instead of one-hot label maps (ablation) | Ours full · COCO-Stuff (retrieving) | FID: 112.83 -> 91.91 | Improved |
| add spatial and channel attention instead of one-hot label maps (ablation) | Ours full · COCO-Stuff (retrieving) | PSNR: 11.96 -> 16.65 | Improved |
| add feature masking (ablation) | Ours full · COCO-Stuff (retrieving) | FID: 100.01 -> 91.91 | Improved |
| add feature masking (ablation) | Ours full · COCO-Stuff (retrieving) | PSNR: 16.24 -> 16.65 | Improved |
| add global-level self-reconstruction and increased patch size (ablation) | Ours full · COCO-Stuff (retrieving) | FID: 96.66 -> 91.91 | Improved |
| add global-level self-reconstruction and increased patch size (ablation) | Ours full · COCO-Stuff (mirroring) | FID: 85.92 -> 76.75 | Improved |
| add feature masking (ablation) | Ours full · COCO-Stuff (mirroring) | FID: 90.52 -> 76.75 | Improved |
| add MSCA module instead of global average pooling (ablation) | Ours full · COCO-Stuff (mirroring) | FID: 89.15 -> 76.75 | Improved |
| add self-supervision (GAP model with self-supervision vs SPADE-VAE self-reconstruction) (ablation) | Ours full · COCO-Stuff (retrieving) | FID: 102.68 -> 101.74 | Improved |
| add self-supervision (GAP model vs SPADE-VAE) in user study (ablation) | Ours full · COCO-Stuff user study | choice rate (%): 15.6 -> 29.3 | Improved |
Closest papers by meaning
- Learning to Manipulate Artistic Images (AAAI 2024 · similarity 0.87)
- Compositional Text-to-Image Synthesis with Attention Map Control of Diffusion Models (AAAI 2024 · similarity 0.87)
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 2024 · similarity 0.86)
- Unleashing In-context Learning of Autoregressive Models for Few-shot Image Manipulation (CVPR 2025 · similarity 0.86)
- Self-Cross Diffusion Guidance for Text-to-Image Synthesis of Similar Subjects (CVPR 2025 · similarity 0.86)
- Grounded Text-to-Image Synthesis with Attention Refocusing (CVPR 2024 · 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.
