Research map / Image generation / GANs
Unsupervised Segmentation Incorporating Shape Prior via Generative Adversarial Networks
Dahye Kim, Byung-Woo Hong
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 shape prior GAN on top of intrinsic image segmentation (ablation) | Intrinsic segmentation · LSUN car | IoU: 0.3250 -> 0.6340 | Improved |
| add shape prior GAN on top of intrinsic image segmentation (ablation) | Intrinsic segmentation · LSUN boat | IoU: 0.2348 -> 0.4901 | Improved |
| add shape prior GAN on top of intrinsic image segmentation (ablation) | Intrinsic segmentation · LSUN airplane | IoU: 0.2307 -> 0.4714 | Improved |
| add shape prior GAN on top of intrinsic image segmentation (ablation) | Intrinsic segmentation · LSUN chair | IoU: 0.3683 -> 0.4776 | Improved |
| add intrinsic image decomposition before segmentation with shape prior (ablation) | Original + shape · LSUN car | IoU: 0.6303 -> 0.6340 | Improved |
| add intrinsic image decomposition before segmentation with shape prior (ablation) | Original + shape · LSUN boat | IoU: 0.4756 -> 0.4901 | Improved |
| add intrinsic image decomposition before segmentation with shape prior (ablation) | Original + shape · LSUN chair | IoU: 0.4824 -> 0.4776 | Hurt |
| add intrinsic image decomposition without shape prior (ablation) | Original only · LSUN car | IoU: 0.3126 -> 0.3250 | Improved |
| add intrinsic image decomposition without shape prior (ablation) | Original only · LSUN boat | IoU: 0.2002 -> 0.2348 | Improved |
| add shape prior and intrinsic decomposition vs strongest baseline GrabCut | Full model · LSUN chair | IoU: 0.5127 -> 0.4776 | Hurt |
| add intrinsic decomposition on heavy occlusion with strong bias (ablation) | Original + shape · Synthetic squares (occlusion 80%, bias std 0.4) | IoU: 0.3911 -> 0.7649 | Improved |
| add shape prior on intrinsic images with heavy occlusion (ablation) | Intrinsic segmentation · Synthetic squares (occlusion 80%, bias std 0.4) | IoU: 0.3168 -> 0.7649 | Improved |
Closest papers by meaning
- Object Segmentation Without Labels with Large-Scale Generative Models (ICML 2021 · similarity 0.89)
- Information-Theoretic Segmentation by Inpainting Error Maximization (CVPR 2021 · similarity 0.88)
- Finding an Unsupervised Image Segmenter in each of your Deep Generative Models (ICLR 2022 · similarity 0.88)
- GANSeg: Learning To Segment by Unsupervised Hierarchical Image Generation (CVPR 2022 · similarity 0.87)
- Factorized Diffusion Architectures for Unsupervised Image Generation and Segmentation (NeurIPS 2024 · similarity 0.86)
- ASAM: Boosting Segment Anything Model with Adversarial Tuning (CVPR 2024 · 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.
