Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Meta-Learning of Figure-Ground Segmentation via Imitating Visual Effects
Ding-Jie Chen, Jui-Ting Chien, Hwann-Tzong Chen, Tyng-Luh Liu
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 |
|---|---|---|---|
| use VEGAN instead of the strongest listed unsupervised baseline, CA | figure-ground segmentation · MSRA500 | mean IoU: 0.67 -> 0.76 | Improved |
| use VEGAN instead of the strongest listed unsupervised baseline, MST | figure-ground segmentation · GC50 | mean IoU: 0.60 -> 0.58 | Hurt |
| use VEGAN instead of the strongest listed unsupervised baseline, CA | figure-ground segmentation · ECSSD | mean IoU: 0.50 -> 0.58 | Improved |
| use VEGAN instead of the strongest listed unsupervised baseline, CA | figure-ground segmentation · Flower17 | mean IoU: 0.72 -> 0.72 | No clear effect |
| use VEGAN instead of the strongest listed unsupervised baseline, CA | figure-ground segmentation · Flower102 | mean IoU: 0.75 -> 0.81 | Improved |
| use VEGAN instead of the strongest listed unsupervised baseline, MST | figure-ground segmentation · CUB200 | mean IoU: 0.54 -> 0.52 | Hurt |
| replace the base generator with pre-trained ResNet (ablation) | VEGAN B4, black background · MSRA500 | mean IoU: 0.76 -> 0.73 | Hurt |
| add skip-layers and bilinear upsampling while retaining patchGAN (ablation) | VEGAN B4, black background · MSRA500 | mean IoU: 0.76 -> 0.70 | Hurt |
| use full-image discriminator with skip-layers and bilinear upsampling rather than patchGAN without those modifications (ablation) | VEGAN, black background · MSRA500 | mean IoU: 0.67 -> 0.76 | Improved |
| replace the base generator with pre-trained ResNet (ablation) | VEGAN C4, color selectivo · MSRA500 | mean IoU: 0.75 -> 0.73 | Hurt |
| add skip-layers and bilinear upsampling while retaining patchGAN (ablation) | VEGAN C4, color selectivo · MSRA500 | mean IoU: 0.75 -> 0.74 | Hurt |
| replace the base generator with pre-trained ResNet (ablation) | VEGAN D4, defocus/Bokeh · MSRA500 | mean IoU: 0.73 -> 0.66 | Hurt |
Closest papers by meaning
- Early Semantic Grounding in Image Editing Models for Zero-Shot Referring Image Segmentation (NeurIPS 2026 · similarity 0.86)
- A Generative Approach at the Instance-Level for Image Segmentation Under Limited Training Data Conditions (Student Abstract) (AAAI 2025 · similarity 0.85)
- Prompting Diffusion Models for Zero-Shot Instance Segmentation (NeurIPS 2026 · similarity 0.85)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.85)
- UGround: Towards Unified Visual Grounding with Unrolled Transformers (ICML 2026 · similarity 0.85)
- Segment Anything without Supervision (NeurIPS 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.
