Research map / Image generation / GANs
Noise Robust Generative Adversarial Networks
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 |
|---|---|---|---|
| Use a transformation-constrained noise generator rather than an unconstrained one (ablation) | SI-NR-GAN-II · CIFAR-10, additive Gaussian fixed (A) | FID: 40.7 -> 29.8 | Helped |
| Add rotation as the only transformation (ablation) | SI-NR-GAN-II · CIFAR-10, additive Gaussian fixed (A) | FID: 29.8 -> 28.3 | Helped |
| Add color inversion as the only transformation (ablation) | SI-NR-GAN-II · CIFAR-10, local Gaussian fixed (C) | FID: 23.1 -> 22.3 | Helped |
| Use all three transformations instead of rotation alone (ablation) | SI-NR-GAN-II · CIFAR-10, additive Gaussian variable (B) | FID: 29.8 -> 29.7 | No clear effect |
| Add vertical and horizontal blur filtering (Blurvh) (ablation) | SD-NR-GAN-II · CIFAR-10, Poisson fixed (M) | FID: 48.5 -> 34.0 | Helped |
| Replace Blurvh with standard blur (ablation) | SD-NR-GAN-II · CIFAR-10, Poisson fixed (M) | FID: 34.0 -> 57.8 | Hurt |
| Remove the noise-generator constraint (ablation) | SD-NR-GAN-II · CIFAR-10, clean | FID: 19.3 -> 57.0 | Hurt |
| Use diversity-sensitive regularization with appropriate strength (ablation) | NR-GANs · CIFAR-10, noisy images (A, B, I, M) | FID | Helped |
| Use diversity-sensitive regularization at λ ∈ [0.01, 0.02] on clean data (ablation) | NR-GANs · CIFAR-10, clean | FID: within 1.0 | Hurt |
| Use strong distribution constraints rather than weak constraints | NR-GANs · Noise conditions matching the assumed noise model | FID | Helped |
| Use weak transformation constraints instead of a fixed noise-type assumption | SI-NR-GAN-II vs SI-NR-GAN-I · CIFAR-10, out-of-assumption noise including uniform and Brown Gaussian | FID: 31.6 vs 40.1; 32.2 vs 163.4 | Helped |
| Use an explicitly specified signal-noise relationship | SD-NR-GAN-I · CIFAR-10, signal-dependent noise within or near assumptions | FID: 22.5 vs 24.4 | Helped |
Closest papers by meaning
- Blur, Noise, and Compression Robust Generative Adversarial Networks (CVPR 2021 · similarity 0.93)
- Label-Noise Robust Generative Adversarial Networks (CVPR 2019 · similarity 0.92)
- Learning to Generate Realistic Noisy Images via Pixel-level Noise-aware Adversarial Training (NeurIPS 2021 · similarity 0.92)
- sRGB Real Noise Modeling via Noise-Aware Sampling with Normalizing Flows (ICLR 2024 · similarity 0.90)
- First Line of Defense: A Robust First Layer Mitigates Adversarial Attacks (AAAI 2025 · similarity 0.87)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · 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.