Research map / Image generation / Editing and style transfer
SHUNIT: Style Harmonization for Unpaired Image-to-Image Translation
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 |
|---|---|---|---|
| Add component style and class-wise alpha to memory style (ablation) | SHUNIT · Cityscapes clear → ACDC snow | mIoU: 26.33 -> 44.35 | Helped |
| Add component style and class-wise alpha to memory style (ablation) | SHUNIT · Cityscapes clear → ACDC rain | mIoU: 39.06 -> 42.77 | Helped |
| Add component style while omitting class-wise alpha (ablation) | SHUNIT · Cityscapes clear → ACDC rain | mIoU: 39.06 -> 38.97 | No clear effect |
| Add class-wise alpha to the style harmonization layer (ablation) | SHUNIT · Cityscapes clear → ACDC snow | mIoU: 37.20 -> 44.35 | Helped |
| Add class-wise alpha to the style harmonization layer (ablation) | SHUNIT · Cityscapes clear → ACDC rain | mIoU: 38.97 -> 42.77 | Helped |
| Use content and style contrastive losses jointly (ablation) | SHUNIT · Cityscapes clear → ACDC snow | mIoU: 32.43 -> 44.35 | Helped |
| Use content and style contrastive losses jointly (ablation) | SHUNIT · Cityscapes clear → ACDC rain | mIoU: 39.47 -> 42.77 | Helped |
| Train style memory with backpropagation instead of updating (ablation) | SHUNIT · Cityscapes clear → ACDC snow | mIoU: 36.70 -> 44.35 | Helped |
| Train style memory with backpropagation instead of updating (ablation) | SHUNIT · Cityscapes clear → ACDC rain | mIoU: 38.21 -> 42.77 | Helped |
| Use contrastive rather than L1 content and style losses (ablation) | SHUNIT · Cityscapes clear → ACDC snow | mIoU: 38.98 -> 44.35 | Helped |
| Use contrastive rather than L1 content and style losses (ablation) | SHUNIT · Cityscapes clear → ACDC rain | cFID: 7.72 -> 8.47 | Hurt |
| Add semantic label input to the content encoder (ablation) | SHUNIT · Cityscapes clear → ACDC snow | mIoU: 44.11 -> 44.35 | No clear effect |
Closest papers by meaning
- Towards Instance-Level Image-To-Image Translation (CVPR 2019 · similarity 0.88)
- Domain Adaptive Image-to-Image Translation (CVPR 2020 · similarity 0.87)
- Style-Guided and Disentangled Representation for Robust Image-to-Image Translation (AAAI 2022 · similarity 0.87)
- Image-to-Image Translation via Hierarchical Style Disentanglement (CVPR 2021 · similarity 0.87)
- Encoding in Style: A StyleGAN Encoder for Image-to-Image Translation (CVPR 2021 · similarity 0.86)
- Smoothing the Disentangled Latent Style Space for Unsupervised Image-to-Image Translation (CVPR 2021 · 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.