Research map / Image generation / GANs
ContraGAN: Contrastive Learning for Conditional Image Generation
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 BigGAN class conditioning with ContraGAN 2C loss | BigGAN · CIFAR10 | FID: 10.739±0.016 -> 10.597±0.273 | Helped |
| replace BigGAN class conditioning with ContraGAN 2C loss | BigGAN · Tiny ImageNet | FID: +7.3% | Helped |
| replace BigGAN class conditioning with ContraGAN 2C loss | BigGAN · ImageNet | FID: 21.072 -> 19.443 | Helped |
| use ContraGAN conditioning instead of ProjGAN conditioning | ResGAN · CIFAR10 | FID: 11.025± 0.177 -> 11.334±0.126 | Hurt |
| combine data-to-class and data-to-data relations in 2C loss rather than Eq.7 loss | ContraGAN · CIFAR10 | FID: 10.886±0.072 -> 10.597±0.273 | Helped |
| combine data-to-class and data-to-data relations in 2C loss rather than Eq.7 loss | ContraGAN · Tiny ImageNet | FID: 33.488±1.006 -> 32.720±1.084 | Helped |
| add 2C loss (ablation) | BigGAN · Tiny ImageNet, batch 256 | FID: 40.981 -> 32.094 | Helped |
| add 2C loss (ablation) | BigGAN · Tiny ImageNet, batch 1024 | FID: 34.090 -> 30.286 | Helped |
| add augmented positive samples (APS) to 2C loss (ablation) | BigGAN with 2C loss · Tiny ImageNet, batch 256 | FID: 32.094 -> 33.151 | Hurt |
| add consistency regularization (CR) to 2C loss (ablation) | BigGAN with 2C loss · Tiny ImageNet, batch 256 | FID: 32.094 -> 28.631 | Helped |
| add consistency regularization (CR) to 2C loss (ablation) | BigGAN with 2C loss · Tiny ImageNet, batch 1024 | FID: 30.286 -> 27.018 | Helped |
| increase batch size from 256 to 1024 with no regularization (ablation) | BigGAN · Tiny ImageNet | FID: 40.981 -> 34.090 | Helped |
Closest papers by meaning
- CcGAN: Continuous Conditional Generative Adversarial Networks for Image Generation (ICLR 2021 · similarity 0.90)
- High-Fidelity Image Generation With Fewer Labels (ICML 2019 · similarity 0.89)
- Robust Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.89)
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.89)
- Turning Waste into Wealth: Leveraging Low-Quality Samples for Enhancing Continuous Conditional Generative Adversarial Networks (AAAI 2024 · similarity 0.88)
- Attack Deterministic Conditional Image Generative Models for Diverse and Controllable Generation (AAAI 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.