Research map / Image generation / Image restoration
Deep Adversarial Decomposition: A Unified Framework for Separating Superimposed Images
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 crossroad l1 loss rather than standard l1 loss (ablation) | Deep Adversarial Decomposition · Stanford-Dogs | PSNR: 16.26 -> 22.74 | Helped |
| Use crossroad l1 loss rather than standard l1 loss (ablation) | Deep Adversarial Decomposition · Stanford-Dogs | SSIM: 0.6768 -> 0.7782 | Helped |
| Add Separation-Critic instead of using cross-l1 alone (ablation) | Deep Adversarial Decomposition · Stanford-Dogs + VGG-Flowers with overexposure and noise | PSNR: 17.34 -> 18.29 | Helped |
| Add Separation-Critic instead of using cross-l1 alone (ablation) | Deep Adversarial Decomposition · Stanford-Dogs + VGG-Flowers with overexposure and noise | SSIM: 0.6693 -> 0.7231 | Helped |
| Add adversarial perceptual loss to the Separation-Critic setup (ablation) | Deep Adversarial Decomposition · Stanford-Dogs + VGG-Flowers with overexposure and noise | SSIM: 0.7231 -> 0.6938 | Hurt |
| Add adversarial perceptual loss to the Separation-Critic setup (ablation) | Deep Adversarial Decomposition · Stanford-Dogs + VGG-Flowers with overexposure and noise | PI: 22.16 -> 21.97 | Helped |
| Use crossroad l1 instead of standard l1 (ablation) | Deep Adversarial Decomposition · Stanford-Dogs | PI: 23.34 -> 21.92 | Hurt |
| Use Deep Adversarial Decomposition instead of Double-DIP | single superimposed image separation · Stanford-Dogs + VGG-Flowers | PSNR: 14.70 -> 25.51 | Helped |
| Use Deep Adversarial Decomposition instead of Double-DIP | single superimposed image separation · LSUN Classroom + LSUN Church | PSNR: 13.83 -> 26.32 | Helped |
| Use Deep Adversarial Decomposition instead of PReNet | image deraining · Rain100H | PSNR: 29.46 -> 30.85 | Helped |
| Use Deep Adversarial Decomposition instead of RESCAN | image deraining · Rain800 | PSNR: 24.09 -> 24.49 | Helped |
| Use Deep Adversarial Decomposition instead of RmNet | reflection removal · Reflection Removal Dataset [59], focused set | PSNR: 21.064 -> 22.809 | Helped |
Closest papers by meaning
- Single Image Reflection Separation With Perceptual Losses (CVPR 2018 · similarity 0.89)
- Separate in Latent Space: Unsupervised Single Image Layer Separation (AAAI 2020 · similarity 0.89)
- Diffusion-based Adversarial Purification from the Perspective of the Frequency Domain (ICML 2025 · similarity 0.84)
- Divide and Conquer: Heterogeneous Noise Integration for Diffusion-based Adversarial Purification (CVPR 2025 · similarity 0.83)
- Adversarial Diffusion Compression for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.83)
- Instant Adversarial Purification with Adversarial Consistency Distillation (CVPR 2025 · similarity 0.83)
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.