Research map / Image generation / Autoregressive and latent generation
Local Class-Specific and Global Image-Level Generative Adversarial Networks for Semantic-Guided Scene Generation
Check 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 |
|---|---|---|---|
| combine global and class-specific local generation using addition (ablation) | Global generator on Cityscapes · Cityscapes | mIoU: 62.3 -> 64.6 | Helped |
| combine global and class-specific local generation using addition (ablation) | Global generator on Cityscapes · Cityscapes | FID: 71.8 -> 66.1 | Helped |
| use convolution rather than addition to combine the local generator outputs (ablation) | Global + Local generation on Cityscapes · Cityscapes | mIoU: 64.6 -> 65.8 | Helped |
| use convolution rather than addition to combine the local generator outputs (ablation) | Global + Local generation on Cityscapes · Cityscapes | FID: 66.1 -> 65.6 | Helped |
| add class-specific discriminative feature learning loss (ablation) | Global + Local generation with convolution on Cityscapes · Cityscapes | mIoU: 65.8 -> 67.0 | Helped |
| add class-specific discriminative feature learning loss (ablation) | Global + Local generation with convolution on Cityscapes · Cityscapes | FID: 65.6 -> 61.3 | Helped |
| add learned pixel-level weight-map fusion (ablation) | Global + Local generation with class-discriminative loss on Cityscapes · Cityscapes | mIoU: 67.0 -> 68.4 | Helped |
| add learned pixel-level weight-map fusion (ablation) | Global + Local generation with class-discriminative loss on Cityscapes · Cityscapes | FID: 61.3 -> 57.7 | Helped |
| use LGGAN rather than SelectionGAN | Cross-view image translation · Dayton a2g | Top-1 accuracy (%): 42.11 -> 48.17 | Helped |
| use LGGAN rather than SelectionGAN | Cross-view image translation · CVUSA a2g | Top-1 accuracy (%): 41.52 -> 44.75 | Helped |
| use LGGAN rather than SelectionGAN | Cross-view image translation · Dayton a2g | SSIM: 0.5938 -> 0.5457 | Hurt |
| use LGGAN rather than SelectionGAN | Cross-view image translation · CVUSA a2g | SSIM: 0.5323 -> 0.5238 | Hurt |
Closest papers by meaning
- Semantic Palette: Guiding Scene Generation With Class Proportions (CVPR 2021 · similarity 0.89)
- Semantic Pyramid for Image Generation (CVPR 2020 · similarity 0.89)
- Adversarially Tuned Scene Generation (CVPR 2017 · similarity 0.88)
- KE-GAN: Knowledge Embedded Generative Adversarial Networks for Semi-Supervised Scene Parsing (CVPR 2019 · similarity 0.88)
- A Layer-Based Sequential Framework for Scene Generation with GANs (AAAI 2019 · similarity 0.88)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.87)
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.