Research map / Detection and segmentation / Open-set and few-shot detection
Promising or Elusive? Unsupervised Object Segmentation from Real-world Single Images
Yafei YANG, Bo Yang
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 |
|---|---|---|---|
| replace each object's pixels with its average color, removing object color gradients (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-C, ScanNet-C, and COCO-C | object segmentation performance | Improved |
| replace object masks with convex shapes, reducing object shape concavity (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-S, ScanNet-S, and COCO-S | object segmentation performance | Improved |
| remove object color gradients and shape concavity together (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-C+S, ScanNet-C+S, and COCO-C+S | object segmentation performance | Improved |
| replace object textures with distinctive textures to reduce inter-object color similarity (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-T, ScanNet-T, and COCO-T | object segmentation performance | Improved |
| normalize object sizes within each image to reduce inter-object shape variation (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-U, ScanNet-U, and COCO-U | object segmentation performance | Improved |
| reduce inter-object color similarity and shape variation together (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-T+U, ScanNet-T+U, and COCO-T+U | object segmentation performance | Improved |
| remove both object-level factors and one scene-level factor (ablation) | IODINE and SlotAtt · YCB, ScanNet, and COCO ablated datasets | object segmentation performance | Improved |
| remove all four appearance and geometry complexity factors (ablation) | AIR, MONet, IODINE, and SlotAtt · YCB-C+S+T+U, ScanNet-C+S+T+U, and COCO-C+S+T+U | object segmentation performance | Improved |
| remove inter-object shape variation (ablation) | AIR · Real-world ablated datasets | object segmentation performance | Improved |
| remove object color gradients and inter-object color similarity (ablation) | MONet · Real-world ablated datasets | object segmentation performance | Improved |
| remove object shape concavity or inter-object shape variation (ablation) | MONet · Real-world ablated datasets | object segmentation performance | No clear effect |
| remove object color gradients (ablation) | IODINE · Real-world ablated datasets | object segmentation performance | Improved |
Closest papers by meaning
- Benchmarking and Analysis of Unsupervised Object Segmentation from Real-World Single Images (IJCV 2024 · similarity 0.99)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.89)
- Unsupervised Object Segmentation in Video by Efficient Selection of Highly Probable Positive Features (ICCV 2017 · similarity 0.89)
- Unsupervised Part Segmentation Through Disentangling Appearance and Shape (CVPR 2021 · similarity 0.88)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.87)
- Segment Anything without Supervision (NeurIPS 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.
