Research map / Image generation / GANs
Spatial Frequency Bias in Convolutional 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 |
|---|---|---|---|
| Shift the dataset frequencies toward high frequencies | WGAN-GP · SCelebA and SBedrooms | FID: 20.97 sd 0.70 -> 328.72 sd 9.70; 55.14 sd 1.29 -> 283.02 sd 7.06 | Hurt |
| Shift the dataset frequencies toward high frequencies | PG-GAN · SCelebA and SBedrooms | FID: 8.66 sd 0.41 -> 23.12 sd 2.08; 36.65 sd 0.97 -> 69.03 sd 10.28 | Hurt |
| Shift the dataset frequencies toward high frequencies | StyleGAN2 · SCelebA and SBedrooms | FID: 6.08 sd 0.11 -> 343.57 sd 53.59; 22.49 sd 2.00 -> 260.84 sd 4.03 | Hurt |
| Add a frequency-shifted generator targeting (1/2, 1/2) | WGAN-GP · SCelebA | FID: 328.72 sd 9.70 -> 20.70 sd 0.44 | Helped |
| Add a frequency-shifted generator targeting (1/2, 1/2) | WGAN-GP · SBedrooms | FID: 283.02 sd 7.06 -> 59.81 sd 1.64 | Helped |
| Add a frequency-shifted generator targeting (1/2, 1/2) | PG-GAN · SCelebA | FID: 23.12 sd 2.08 -> 17.91 sd 0.74 | Helped |
| Add a frequency-shifted generator targeting (1/2, 1/2) | PG-GAN · SBedrooms | FID: 69.03 sd 10.28 -> 54.64 sd 0.26 | Helped |
| Add a frequency-shifted generator targeting (1/2, 1/2) | StyleGAN2 · SCelebA | FID: 343.57 sd 53.59 -> 7.17 sd 0.07 | Helped |
| Add a frequency-shifted generator targeting (1/2, 1/2) | StyleGAN2 · SBedrooms | FID: 260.84 sd 4.03 -> 67.85 sd 2.38 | Helped |
Closest papers by meaning
- On the Frequency Bias of Generative Models (NeurIPS 2021 · similarity 0.93)
- FreGAN: Exploiting Frequency Components for Training GANs under Limited Data (NeurIPS 2022 · similarity 0.88)
- Fourier Spectrum Discrepancies in Deep Network Generated Images (NeurIPS 2020 · similarity 0.88)
- Leveraging Frequency Analysis for Deep Fake Image Recognition (ICML 2020 · similarity 0.87)
- SSD-GAN: Measuring the Realness in the Spatial and Spectral Domains (AAAI 2021 · similarity 0.87)
- Spectrum Translation for Refinement of Image Generation (STIG) Based on Contrastive Learning and Spectral Filter Profile (AAAI 2024 · similarity 0.85)
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.