Research map / Image generation / GANs
GlowGAN: Unsupervised Learning of HDR Images from LDR Images in the Wild
Chao Wang, Ana Serrano, Xingang Pan, Bin Chen, Karol Myszkowski, Hans-Peter Seidel, Christian Theobalt, Thomas Leimkühler
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| learn an HDR generator through a camera model with random exposure instead of vanilla GAN | StyleGAN-XL · Landscapes | DR50: 8.0 -> 15.4 | Improved |
| increase exposure variance sigma_e^2 from 1.0 to 5.0 (ablation) | GlowGAN · Landscapes | FID and DR50: FID 3.61 -> 4.00; DR50 15.4 -> 16.5 | Mixed |
| use stochastic CRF sampling instead of fixed CRF (ablation) | GlowGAN · Landscapes | FID: 3.89 -> 3.61 | Improved |
| use stochastic CRF sampling instead of fixed CRF (ablation) | GlowGAN · Lightning | FID: 3.40 -> 3.29 | Improved |
| switch backbone to StyleGAN2-ADA (ablation) | GlowGAN · Landscapes | DR90: 8.0 -> 22.9 | Improved |
| apply GlowGAN dynamic range training instead of vanilla StyleGAN2-ADA (ablation) | StyleGAN2-ADA · Landscapes | FID: 9.2 -> 10.02 | Hurt |
| use unsupervised GlowGAN for inverse tone mapping versus supervised methods | inverse tone mapping · 100 LDR images, non-reference | PU21-PIQE: 33.3 -> 31.8 (vs MaskHDR) | Improved |
| use unsupervised GlowGAN for inverse tone mapping versus supervised methods | inverse tone mapping · 62 HDR test images | HDR-VDP3: 7.60 (MaskHDR) vs 7.44 | No clear effect |
| use GlowGAN versus six supervised ITM methods in a user study | inverse tone mapping · user study, 24 participants | preference ranking: ranked significantly higher than all others (p < 0.001) | Improved |
Closest papers by meaning
- A Unified HDR Imaging Method With Pixel and Patch Level (CVPR 2023 · similarity 0.89)
- Zero-Shot Structure-Preserving Diffusion Model for High Dynamic Range Tone Mapping (CVPR 2024 · similarity 0.88)
- High Dynamic Range Novel View Synthesis with Single Exposure (ICML 2025 · similarity 0.87)
- GaussHDR: High Dynamic Range Gaussian Splatting via Learning Unified 3D and 2D Local Tone Mapping (CVPR 2025 · similarity 0.87)
- Generating Content for HDR Deghosting from Frequency View (CVPR 2024 · similarity 0.87)
- LEDiff: Latent Exposure Diffusion for HDR Generation (CVPR 2025 · similarity 0.86)
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.
