Research map / Image generation / GANs
OASIS: Only Adversarial Supervision for Semantic Image Synthesis
Vadim Sushko, Edgar Schönfeld, Dan Zhang, Jüergen Gall, Bernt Schiele, Anna Khoreva
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 |
|---|---|---|---|
| replace SPADE+ discriminator with the OASIS segmentation discriminator (ablation) | SPADE+ · ADE20K | FID and mIoU: improves FID and mIoU by more than 30 points | Improved |
| use lighter OASIS generator (remove first residual block) (ablation) | SPADE+ with OASIS D · ADE20K | FID and mIoU: negligible degradation of 0.3 in FID and 0.5 in mIoU | No clear effect |
| add LabelMix consistency regularization (ablation) | OASIS · Cityscapes | FID: 51.5 -> 47.7 | Improved |
| add LabelMix consistency regularization (ablation) | OASIS · Cityscapes | mIoU: 66.3 -> 69.3 | Improved |
| use CutMix instead of LabelMix for consistency regularization (ablation) | OASIS · Cityscapes | FID: 47.7 -> 52.1 | Hurt |
| add 3D noise (ablation) | OASIS · ADE20K | MS-SSIM (diversity): 1.0 -> 0.63 | Improved |
| add 3D noise (ablation) | OASIS · COCO-Stuff | mIoU: 45.5 -> 44.1 | Hurt |
| add VGG perceptual loss (ablation) | OASIS · Cityscapes | FID: 47.7 -> 46.1 | Improved |
| add VGG perceptual loss (ablation) | OASIS · ADE20K | FID: 28.3 -> 31.6 | Hurt |
| add VGG perceptual loss (ablation) | OASIS · ADE20K | diversity (MS-SSIM): 0.65 -> 0.88 | Hurt |
| add feature matching loss (ablation) | SPADE+ (no VGG) · ADE20K | FID: 60.7 -> 49.7 | Improved |
| add feature matching loss (ablation) | OASIS (no VGG) · Cityscapes | FID: 47.7 -> 48.5 | Hurt |
Later papers that compared with this method
9 later accepted papers report OASIS as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.
- Stochastic Conditional Diffusion Models for Robust Semantic Image Synthesis (ICML 2024 · better on 2 of 3 shared results)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · better on 2 of 2 shared results)
- Edge Guided GANs With Multi-Scale Contrastive Learning for Semantic Image Synthesis (TPAMI 2023 · better on 8 of 8 shared results)
- Freestyle Layout-to-Image Synthesis (CVPR 2023 · better on 2 of 4 shared results)
- Semantic Image Synthesis with Unconditional Generator (NeurIPS 2023 · better on 2 of 3 shared results)
- Zero-Shot Spatial Layout Conditioning for Text-to-Image Diffusion Models (ICCV 2023 · better on 1 of 3 shared results)
- Edge Guided GANs with Contrastive Learning for Semantic Image Synthesis (ICLR 2023 · better on 2 of 2 shared results)
- Retrieval-Based Spatially Adaptive Normalization for Semantic Image Synthesis (CVPR 2022 · better on 4 of 7 shared results)
1 more not listed.
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- You Only Need Adversarial Supervision for Semantic Image Synthesis (ICLR 2021 · similarity 0.98)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.91)
- Linear Semantics in Generative Adversarial Networks (CVPR 2021 · similarity 0.91)
- 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)
- Edge Guided GANs With Multi-Scale Contrastive Learning for Semantic Image Synthesis (TPAMI 2023 · 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.
