Research map / Image generation / GANs
You Only Need Adversarial Supervision for Semantic Image Synthesis
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 |
|---|---|---|---|
| Replace the classification discriminator with a segmentation-based OASIS discriminator (ablation) | SPADE+ generator · ADE20K and Cityscapes | FID: 60.7 -> 29.3; 61.4 -> 54.1 | Helped |
| Add semantic-boundary LabelMix consistency regularization (ablation) | OASIS · Cityscapes | FID: 51.5 -> 47.7 | Helped |
| Add semantic-boundary LabelMix consistency regularization (ablation) | OASIS · Cityscapes | mIoU: 66.3 -> 69.3 | Helped |
| Use CutMix consistency regularization (ablation) | OASIS · Cityscapes | FID: 51.5 -> 52.1 | Hurt |
| Add 3D noise for multimodal synthesis (ablation) | OASIS · ADE20K and Cityscapes | MS-SSIM: 1.0 -> 0.63; 1.0 -> 0.62 | Helped |
| Add 3D noise for multimodal synthesis (ablation) | OASIS · Cityscapes | mIoU: 69.3 -> 66.3 | Hurt |
| Add VGG perceptual loss (ablation) | OASIS with 3D noise and LabelMix · ADE20K | FID: 28.3 -> 31.6 | Hurt |
| Add VGG perceptual loss (ablation) | OASIS with 3D noise and LabelMix · Cityscapes | MS-SSIM: 0.64 -> 0.84 | Hurt |
| Replace the full OASIS generator with its lighter version (ablation) | OASIS · ADE20K | FID: 0.3 | No clear effect |
| Add feature matching loss (ablation) | OASIS without VGG · Cityscapes | FID: 47.7 -> 48.5 | Hurt |
| Add a second discriminator at scale 0.5 (ablation) | OASIS · Cityscapes | FID: 47.7 -> 48.7 | Hurt |
| Use class-balanced N+1 loss instead of unbalanced N+1 loss (ablation) | OASIS discriminator · ADE20K rare semantic classes | mIoU: 36.2 -> 41.5; 31.2 -> 39.7 | Helped |
Closest papers by meaning
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.92)
- Linear Semantics in Generative Adversarial Networks (CVPR 2021 · similarity 0.92)
- Learning to Predict Layout-to-image Conditional Convolutions for Semantic Image Synthesis (NeurIPS 2019 · similarity 0.91)
- Edge Guided GANs with Contrastive Learning for Semantic Image Synthesis (ICLR 2023 · similarity 0.91)
- Few-Shot Semantic Image Synthesis With Class Affinity Transfer (CVPR 2023 · similarity 0.91)
- Conditional Image Synthesis with Auxiliary Classifier GANs (ICML 2017 · similarity 0.90)
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.