Research map / Image generation / Image restoration
Unsupervised Variational Translator for Bridging Image Restoration and High-Level Vision Tasks
Jiawei Wu, Zhi Jin
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 the maximum likelihood (L_mle) term to the cycle loss (ablation) | VaT · RTTS dehazing detection | mAP %: 11.45 -> 42.77 | Improved |
| add cycle-consistency loss to L_mle (ablation) | VaT · RTTS dehazing detection | mAP %: 41.97 -> 42.77 | Improved |
| add gated fusion module (ablation) | VaT · RTTS dehazing detection | mAP %: 42.77 -> 47.49 | Improved |
| add gated fusion module (ablation) | VaT · Low-light CIFAR-10 CLIP | accuracy %: 35.10 -> 41.62 | Improved |
| add Mixup augmentation (ablation) | VaT · RTTS dehazing detection | mAP %: 47.49 -> 49.12 | Improved |
| use uncertainty-guided UMix instead of plain Mixup (ablation) | VaT · Low-light CIFAR-10 CLIP | accuracy %: 43.19 -> 44.27 | Improved |
| use a stronger restoration network (DCP, AODNet, GCANet by PSNR) (ablation) | VaT · VOCh-test | detection mAP | Improved |
| increase translator size (0.21M to 6.10M parameters) (ablation) | VaT · RTTS | detection mAP | Improved |
| apply VaT over unsupervised translation methods (CycleGAN, CUT, MSPC, SRC, SPR) | AODNet + YOLOv3 · RTTS | mAP: 33.51 -> 49.48 | Improved |
| VaT versus supervised AD-YOLO on real hazy data | VaT · RTTS | mAP: 45.47 -> 49.48 | Improved |
| VaT versus supervised AD-YOLO on synthetic hazy data | VaT · VOCh-test | mAP: 84.11 -> 83.59 | No clear effect |
| VaT image quality versus unpaired methods on synthetic hazy data | VaT · VOCh-test | PSNR: 27.871 -> 26.682 | Hurt |
Closest papers by meaning
- Vision-Language Gradient Descent-driven All-in-One Deep Unfolding Networks (CVPR 2025 · similarity 0.87)
- Omni-Kernel Network for Image Restoration (AAAI 2024 · similarity 0.86)
- Sharing Key Semantics in Transformer Makes Efficient Image Restoration (NeurIPS 2024 · similarity 0.86)
- LaVin-DiT: Large Vision Diffusion Transformer (CVPR 2025 · similarity 0.86)
- Boosting Image Restoration via Priors from Pre-trained Models (CVPR 2024 · similarity 0.86)
- Latent Harmony: Synergistic Unified UHD Image Restoration via Latent Space Regularization and Controllable Refinement (NeurIPS 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.
