Research map / Image generation / Image restoration
Blind Image Decomposition
Junlin Han, Weihao Li, Pengfei Fang, Chunyi Sun, Jie Hong, Mohammad Ali Armin, Lars Petersson, Hongdong 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 multi-scale encoder (vs single-scale) (ablation) | BIDeN · Task I (N=4) | Fruit PSNR: 17.26 -> 18.19 | Improved |
| add adversarial loss (ablation) | BIDeN · Task I (N=4) | Acc (AB): 0.566 -> 0.716 | Improved |
| add perceptual loss (ablation) | BIDeN · Task I (N=4) | Fruit PSNR: 16.67 -> 18.19 | Improved |
| add L1/L2 loss (ablation) | BIDeN · Task I (N=4) | Fruit PSNR: 15.56 -> 18.19 | Improved |
| add binary cross-entropy loss for source prediction (ablation) | BIDeN · Task I (N=4) | Acc (AB): 0.0 -> 0.716 | Improved |
| move source prediction branch from discriminator into generator (ablation) | BIDeN · Task I (N=4) | Acc (AB): 0.716 -> 0.634 | Hurt |
| share weights between prediction and separation discriminator branches (ablation) | BIDeN · Task I (N=4) | Acc (AB): 0.520 -> 0.716 | Improved |
| add zeroed loss on unused heads (ablation) | BIDeN · Task I (N=4) | Fruit PSNR: 18.19 -> 15.68 | Hurt |
| train on BID setting rather than conventional decomposition | BIDeN · Task II.B real snow | BRISQUE: 22.64 -> 22.40 | No clear effect |
| train on BID setting rather than conventional decomposition | MPRNet · Task II.B real raindrop | BRISQUE: 29.42 -> 25.86 | Improved |
| increase max number of source components N from 2 to 8 | BIDeN · Task I Fruit PSNR | PSNR: 20.07 -> 16.49 | Hurt |
| increase max number of source components N from 2 to 8 | BIDeN · Task I Acc (All) | Acc: 0.957 -> 0.278 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Restormer — Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022) · better on 2 of 3 shared results
- Double-DIP — "Double-DIP": Unsupervised Image Decomposition via Coupled Deep-Image-Priors (CVPR 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Joint Rain Detection and Removal from a Single Image with Contextualized Deep Networks (TPAMI 2019 · similarity 0.88)
- Not Just Streaks: Towards Ground Truth for Single Image Deraining (ECCV 2022 · similarity 0.88)
- Depth-Aware Blind Image Decomposition for Real-World Adverse Weather Recovery (ECCV 2024 · similarity 0.88)
- Removing Rain From Single Images via a Deep Detail Network (CVPR 2017 · similarity 0.88)
- Deep Joint Rain Detection and Removal From a Single Image (CVPR 2017 · similarity 0.88)
- Spatial Attentive Single-Image Deraining With a High Quality Real Rain Dataset (CVPR 2019 · similarity 0.88)
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.
