Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Wide-Angle Portraits Correction by Multi-Scale Transformer
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 MS-Unet instead of Tan's method | Wide-angle portrait correction · Tan's test set | LineAcc: 66.784 -> 67.209 | Helped |
| Use MS-Unet instead of Tan's method | Wide-angle portrait correction · Google's test set | LineAcc: 64.650 -> 66.098 | Helped |
| Add MSTB to the Swin-Unet baseline (ablation) | MS-Unet · Tan's dataset | LineAcc: 66.514 -> 66.763 | Helped |
| Add SFB to Baseline+MSTB (ablation) | MS-Unet · Tan's dataset | LineAcc: 66.763 -> 66.825 | Helped |
| Add the segmentation surrogate task (ablation) | MS-Unet · Tan's dataset | LineAcc: 66.825 -> 66.871 | Helped |
| Add DRC to the segmentation-task variant (ablation) | MS-Unet · Tan's dataset | LineAcc: 66.871 -> 67.154 | Helped |
| Add RC alone without segmentation or DRC (ablation) | MS-Unet · Tan's dataset | LineAcc: 66.825 -> 66.848 | No clear effect |
| Add segmentation, DRC, and RC (ablation) | MS-Unet · Tan's dataset | LineAcc: 66.825 -> 67.209 | Helped |
| Increase the unlabeled set from 0 to 5000 images (ablation) | Semi-supervised MS-Unet · Tan's dataset | LineAcc: 66.871 -> 67.209 | Helped |
| Increase the unlabeled set from 0 to 5000 images (ablation) | Semi-supervised MS-Unet · Tan's dataset | ShapeAcc: 97.493 -> 97.500 | Helped |
| Train with the semi-supervised scheme rather than fully supervised (ablation) | RefineNet · Wide-angle correction test sets | LineAcc: 66.348 -> 66.569 | Helped |
| Train with the semi-supervised scheme rather than fully supervised (ablation) | UNet · Wide-angle correction test sets | LineAcc: 65.246 -> 66.534 | Helped |
Closest papers by meaning
- MaDCoW: Marginal Distortion Correction for Wide-Angle Photography with Arbitrary Objects (CVPR 2025 · similarity 0.81)
- RetouchFormer: Semi-supervised High-Quality Face Retouching Transformer with Prior-Based Selective Self-Attention (AAAI 2024 · similarity 0.80)
- EFormer: Enhanced Transformer towards Semantic-Contour Features of Foreground for Portraits Matting (CVPR 2024 · similarity 0.79)
- An Image is Worth More Than 16x16 Patches: Exploring Transformers on Individual Pixels (ICLR 2025 · similarity 0.79)
- Tangent Transformers for Composition,Privacy and Removal (ICLR 2024 · similarity 0.78)
- Learning Where to Edit Vision Transformers (NeurIPS 2024 · similarity 0.78)
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.