Research map / Image generation / GANs
Hierarchical Modes Exploring in Generative Adversarial Networks
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 |
|---|---|---|---|
| Add hierarchical mode-exploring regularization | DCGAN · CIFAR-10 | FID: 32.21\pm 0.05 -> 28.84\pm 0.05 | Helped |
| Add hierarchical mode-exploring regularization | Pix2Pix · Facades | LPIPS: 0.005\pm 0.001 -> 0.192\pm 0.001 | Helped |
| Add hierarchical mode-exploring regularization | DRIT · Dog2Cat | FID: 59.05\pm 0.31 -> 28.97\pm 0.54 | Helped |
| Add hierarchical mode-exploring regularization | StackGAN++ · CUB-200-2011 | LPIPS: 0.364\pm 0.005 -> 0.376\pm 0.005 | Helped |
| Set every ERE λ⁽ⁱ⁾ to 1 instead of 0 (ablation) | HMGAN with Pix2Pix · Facades | LPIPS: 0.192\pm 0.001 -> 0.007\pm 0.001 | Mixed |
| Set every ERE λ⁽ⁱ⁾ to 1 instead of 0 (ablation) | HMGAN with Pix2Pix · Maps | LPIPS: 0.205\pm 0.001 -> 0.003\pm 0.001 | Hurt |
| Set every ERE λ⁽ⁱ⁾ to 1 instead of 0 (ablation) | HMGAN with StackGAN++ · CUB-200-2011 | FID: 25.40\pm 2.00 -> 27.00\pm 1.25 | Hurt |
| Set every ERE λ⁽ⁱ⁾ to 0.5 (ablation) | HMGAN with StackGAN++ · CUB-200-2011 | FID: 25.40\pm 2.00 -> 25.55\pm 1.50 | No clear effect |
| Set the target expansion ratio smaller (ablation) | HMGAN · Text-to-image generation | LPIPS diversity at feature level 5 | Helped |
| Use the hierarchical regularizer instead of the output-distance regularizer | HMGAN · Categorical generation | results better: 85\% | Helped |
| Use the hierarchical regularizer instead of the output-distance regularizer | HMGAN · Paired image-to-image generation | results better: 50\% | Mixed |
Closest papers by meaning
- Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis (CVPR 2019 · similarity 0.95)
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.93)
- UniGAN: Reducing Mode Collapse in GANs using a Uniform Generator (NeurIPS 2022 · similarity 0.90)
- Diverse Image Generation via Self-Conditioned GANs (CVPR 2020 · similarity 0.89)
- Robust Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.89)
- 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.