Research map / Image generation / GANs
Boundless: Generative Adversarial Networks for Image Extension
Piotr Teterwak, Aaron Sarna, Dilip Krishnan, Aaron Maschinot, David Belanger, Ce Liu, William T. Freeman
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 semantic (InceptionV3 activation) conditioning to the discriminator (ablation) | Boundless · Places365 top-50 classes, 25% extension | FID: 0.85 -> 0.79 | Improved |
| add semantic conditioning to the discriminator (ablation) | Boundless · Places365 top-50 classes, 50% extension | FID: 5.01 -> 3.46 | Improved |
| add semantic conditioning to the discriminator (ablation) | Boundless · Places365 top-50 classes, 75% extension | FID: 19.17 -> 8.79 | Improved |
| add semantic conditioning to the discriminator (ablation) | Boundless · Places365 top-50 classes, central inpainting | FID: 2.73 -> 2.53 | Improved |
| add semantic conditioning to the discriminator (ablation) | Boundless · Places365 top-50 classes, 75% extension | PSNR: 9.08 -> 8.07 | Hurt |
| replace Boundless conditioning with a perceptual loss | Boundless · Places365 top-50 classes, 75% extension | FID: 8.79 -> 14.15 | Hurt |
| replace Boundless conditioning with feature matching | Boundless · Places365 top-50 classes, 75% extension | FID: 8.79 -> 14.74 | Hurt |
| replace Boundless conditioning with a perceptual loss | Boundless · Places365 top-50 classes, 25% extension | FID: 0.79 -> 0.40 | Improved |
| replace Boundless conditioning with a perceptual loss | Boundless · Places365 top-50 classes, 50% extension | FID: 3.46 -> 2.32 | Improved |
| use Boundless instead of DeepFill with contextual attention and refinement network | Boundless · Places365 top-50 classes, 75% extension | FID: 31.21 -> 8.79 | Improved |
| remove skip connections in the generator (ablation) | Boundless · Places365 qualitative ablation | visual quality | Improved |
| add instance normalization to the generator (ablation) | Boundless · Places365 qualitative ablation | visual quality | Improved |
Closest papers by meaning
- InOut: Diverse Image Outpainting via GAN Inversion (CVPR 2022 · similarity 0.88)
- Constrained Generative Adversarial Networks for Interactive Image Generation (CVPR 2019 · similarity 0.88)
- Navigating the GAN Parameter Space for Semantic Image Editing (CVPR 2021 · similarity 0.87)
- Neural Photo Editing with Introspective Adversarial Networks (ICLR 2017 · similarity 0.87)
- Linear Semantics in Generative Adversarial Networks (CVPR 2021 · similarity 0.87)
- InstantPainting: Expanding GANs for Efficient Text-Conditioned Image Generation Platform (AAAI 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.
