Research map / Image generation / GANs
DivCo: Diverse Conditional Image Synthesis via Contrastive Generative Adversarial Network
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 |
|---|---|---|---|
| Add latent-augmented contrastive loss to BicycleGAN, replacing latent regression | BicycleGAN · Facades | FID: 80.34\pm 1.46 -> 75.96\pm 1.25 | Helped |
| Add latent-augmented contrastive loss to BicycleGAN, replacing latent regression | BicycleGAN · Maps | FID: 130.74\pm 2.63 -> 109.54\pm 4.41 | Helped |
| Replace latent regression with mode seeking loss | BicycleGAN · Facades | LPIPS: 0.2355\pm 0.0067 -> 0.2677 \pm 0.0259 | Helped |
| Replace latent regression with mode seeking loss | BicycleGAN · Facades | FID: 80.34\pm 1.46 -> 78.32\pm 3.32 | Hurt |
| Use latent-augmented contrastive loss with latent-space positives and negatives | DRIT · Dog2Cat, Cat2Dog, Winter2Summer, Summer2Winter | LPIPS | Helped |
| Increase contrastive-loss weight above 3 (ablation) | DivCo on CIFAR10 · CIFAR10 | FID | Hurt |
| Decrease contrastive-loss weight (ablation) | DivCo on CIFAR10 · CIFAR10 | FID | Hurt |
| Increase latent positive-neighborhood radius (ablation) | DivCo on CIFAR10 · CIFAR10 | FID | Hurt |
| Reduce latent positive-neighborhood radius toward zero (ablation) | DivCo on CIFAR10 · CIFAR10 | FID | Hurt |
| Use an unsuitable contrastive temperature (ablation) | DivCo on CIFAR10 · CIFAR10 | FID | Hurt |
Closest papers by meaning
- Conditional Image Synthesis with Auxiliary Classifier GANs (ICML 2017 · similarity 0.88)
- High-Resolution Image Synthesis and Semantic Manipulation With Conditional GANs (CVPR 2018 · similarity 0.86)
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.86)
- Harmonizing Maximum Likelihood with GANs for Multimodal Conditional Generation (ICLR 2019 · similarity 0.86)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.84)
- Attack Deterministic Conditional Image Generative Models for Diverse and Controllable Generation (AAAI 2024 · similarity 0.84)
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.