Research map / Detection and segmentation / Open-set and few-shot detection
Network-Free, Unsupervised Semantic Segmentation With Synthetic Images
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 style-mixing pixel correlation for segmentation | StyleGAN2 images · LSUN-Horse | mIOU: 0.62 -> 0.77 | Helped |
| Use style-mixing pixel correlation for segmentation | StyleGAN2 images · DeepRoom-livingroom sofa | mIOU | Helped |
| Use synthetic masks generated by Ours to train a U-Net | CelebA-Mask-HQ segmentation · CelebA-Mask-HQ test | mIOU: 0.84 -> 0.87 | Helped |
| Set style-mixing cutoff to c=6 rather than c=4 (ablation) | Ours · FFHQ, 500 images | mIOU: 0.66 -> 0.74 | Helped |
| Set style-mixing cutoff to c=8 rather than c=6 (ablation) | Ours · FFHQ, 500 images | mIOU: 0.74 -> 0.76 | Helped |
| Set style-mixing cutoff to c=12 rather than c=8 (ablation) | Ours · FFHQ, 500 images | mIOU: 0.76 -> 0.44 | Hurt |
| Set style-mixing cutoff to c=8 rather than c=4 (ablation) | Ours · FFHQ, 500 images | mIOU: 0.66 -> 0.76 | Helped |
| Use LAB instead of RGB (ablation) | Ours (c=6, k=2) · FFHQ, 500 images | mIOU: 0.74 -> 0.74 | No clear effect |
| Use LAB instead of RGB (ablation) | Ours (c=6, k=4) · FFHQ, 500 images | mIOU: 0.75 -> 0.79 | Helped |
| Use k=4 rather than k=2 clusters (ablation) | Ours (c=6, RGB) · FFHQ, 500 images | mIOU: 0.74 -> 0.75 | Helped |
| Use k=10 rather than k=4 clusters (ablation) | Ours (c=6, RGB) · FFHQ, 500 images | mIOU: 0.75 -> 0.68 | Hurt |
Closest papers by meaning
- Learning Semantic Segmentation From Synthetic Data: A Geometrically Guided Input-Output Adaptation Approach (CVPR 2019 · similarity 0.91)
- Learning From Synthetic Data: Addressing Domain Shift for Semantic Segmentation (CVPR 2018 · similarity 0.90)
- FreeMask: Synthetic Images with Dense Annotations Make Stronger Segmentation Models (NeurIPS 2023 · similarity 0.89)
- Not All Areas Are Equal: Transfer Learning for Semantic Segmentation via Hierarchical Region Selection (CVPR 2019 · similarity 0.89)
- Repurposing GANs for One-Shot Semantic Part Segmentation (CVPR 2021 · similarity 0.88)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · 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.