Research map / Image generation / Image restoration
Unpaired Deep Image Dehazing Using Contrastive Disentanglement Learning
Xiang Chen, Zhentao Fan, Pengpeng Li, Longgang Dai, Caihua Kong, Zhuoran Zheng, Yufeng Huang, Yufeng Li
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 |
|---|---|---|---|
| add adversarial contrastive loss with negative generators (vs conventional contrastive loss) (ablation) | CDD-GAN · Foggy Cityscapes | PSNR: 19.88 -> 20.35 | Improved |
| add adversarial contrastive loss with negative generators (vs conventional contrastive loss) (ablation) | CDD-GAN · Foggy Cityscapes | SSIM: 0.846 -> 0.843 | No clear effect |
| add a contrastive loss to the CycleGAN baseline (ablation) | CycleGAN baseline · Foggy Cityscapes | PSNR: 18.64 -> 19.88 | Improved |
| add diversity loss L_div (ablation) | CDD-GAN with L_ac · Foggy Cityscapes | PSNR: 20.35 -> 20.71 | Improved |
| add diversity loss L_div (ablation) | CDD-GAN with L_ac · Foggy Cityscapes | SSIM: 0.843 -> 0.865 | Improved |
| add total variation loss L_tv (ablation) | CDD-GAN · Foggy Cityscapes | PSNR / SSIM: 20.71 / 0.865 -> 20.79 / 0.871 | Improved |
| add dark channel loss L_dc (ablation) | CDD-GAN · Foggy Cityscapes | PSNR / SSIM: 20.79 / 0.871 -> 20.93 / 0.874 | Improved |
| add dual learning setting S_dual (ablation) | CDD-GAN · Foggy Cityscapes | PSNR / SSIM: 20.72 / 0.863 -> 20.93 / 0.874 | Improved |
| increase number of negatives from 64 to 256 (ablation) | CDD-GAN · Foggy Cityscapes | PSNR / SSIM: 19.84 / 0.859 -> 20.93 / 0.874 | Improved |
| increase number of negatives from 128 to 256 (ablation) | CDD-GAN · Foggy Cityscapes | PSNR / SSIM: 20.52 / 0.868 -> 20.93 / 0.874 | Improved |
| increase number of negatives from 256 to 512 (ablation) | CDD-GAN · Foggy Cityscapes | PSNR / SSIM: 20.93 / 0.874 -> 20.85 / 0.876 | No clear effect |
| use CDD-GAN dehazed output as preprocessing for YOLOv3 detection | CDD-GAN · RTTS (100 images) | mAP (%): 61.35 -> 66.04 | Improved |
Closest papers by meaning
- Towards Perceptual Image Dehazing by Physics-Based Disentanglement and Adversarial Training (AAAI 2018 · similarity 0.91)
- You Only Look Yourself: Unsupervised and Untrained Single Image Dehazing Neural Network (IJCV 2021 · similarity 0.90)
- Contrastive Learning for Compact Single Image Dehazing (CVPR 2021 · similarity 0.89)
- Dehaze-RetinexGAN: Real-World Image Dehazing via Retinex-based Generative Adversarial Network (AAAI 2025 · similarity 0.87)
- Exploiting Diffusion Prior for Real-World Image Dehazing with Unpaired Training (AAAI 2025 · similarity 0.87)
- ODCR: Orthogonal Decoupling Contrastive Regularization for Unpaired Image Dehazing (CVPR 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.
