Research map / Image generation / Editing and style transfer
Separate in Latent Space: Unsupervised Single Image Layer Separation
Read the paperCheck 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 unsupervised layer separation instead of existing unsupervised methods | Synthetic layer separation · Synthetic dataset | MSE: 0.67/0.61 and 0.46/0.37 -> 0.289/0.274 | Helped |
| Use unsupervised layer separation instead of LK18 | Single image layer separation · MIT intrinsic dataset | Average MSE: 0.0197 -> 0.0154 | Helped |
| Use unsupervised layer separation instead of FY18 | Single image layer separation · MIT intrinsic dataset | Total LMSE: 0.020 -> 0.025 | Hurt |
| Use unsupervised layer separation instead of UNIT | Reflection removal · Reflection removal dataset, background | SSIM: 0.738 -> 0.821 | Helped |
| Use unsupervised layer separation instead of LM14 | Reflection removal · Reflection removal dataset, reflection | PSNR: 17.58 -> 19.31 | Helped |
| Add self-supervision (SS) (ablation) | Unsupervised SILS · Synthetic layer separation | MSE | Helped |
| Add cycle consistency (CC) (ablation) | Unsupervised SILS · Synthetic layer separation | MSE | Helped |
| Balance intra-domain and inter-domain distance losses at (λ1+λ2)/λ3=1 instead of 0.1 (ablation) | Self-supervision · Synthetic layer separation | MSE | Helped |
| Set SS and cycle-consistency weights to w1=1 and w2=10 (ablation) | Unsupervised SILS · Synthetic layer separation | MSE | Helped |
Closest papers by meaning
- Deep Adversarial Decomposition: A Unified Framework for Separating Superimposed Images (CVPR 2020 · similarity 0.89)
- "Double-DIP": Unsupervised Image Decomposition via Coupled Deep-Image-Priors (CVPR 2019 · similarity 0.87)
- Single Image Reflection Separation With Perceptual Losses (CVPR 2018 · similarity 0.86)
- Finding an Unsupervised Image Segmenter in each of your Deep Generative Models (ICLR 2022 · similarity 0.85)
- Object Segmentation Without Labels with Large-Scale Generative Models (ICML 2021 · similarity 0.85)
- ReflexSplit: Single Image Reflection Separation via Layer Fusion-Separation (CVPR 2026 · similarity 0.85)
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.