Research map / Detection and segmentation / Open-set and few-shot detection
Benchmarking Semantic Segmentation Models via Appearance and Geometry Attribute Editing
Zijin Yin, Bing Li, Kongming Liang, Han Sun, Zhongjiang He, Zhanyu Ma, Jun Hai Guo
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 short captions instead of long captions for editing (ablation) | Attribute editing pipeline (Stable Diffusion 2) · Appearance editing | CLIP Score: 0.171 -> 0.512 | Improved |
| use short captions instead of long captions for geometry editing (ablation) | Attribute editing pipeline (Stable Diffusion 2) · Geometry editing | FID: 35.38 -> 16.84 | Improved |
| use short captions instead of medium captions for geometry editing (ablation) | Attribute editing pipeline (Stable Diffusion 2) · Geometry editing | HPSv2: 19.38 -> 22.79 | Improved |
| use short captions instead of long captions for appearance editing (ablation) | Attribute editing pipeline (Stable Diffusion 2) · Appearance editing | Masked LPIPS: 0.346 -> 0.192 | Improved |
| use Mask2Former instead of UperNet as surrogate model for noise filtering (ablation) | Noise filtering · PASCAL VOC | CLIP Accuracy: 0.879 -> 0.903 | No clear effect |
| use Mask2Former instead of UperNet as surrogate model for noise filtering (ablation) | Noise filtering · COCO Stuff 164k | CLIP Accuracy: 0.882 -> 0.915 | No clear effect |
| stronger surrogate model filters more samples (ablation) | Noise filtering · PASCAL VOC | Filtered/Total: 14/200 -> 23/200 | Mixed |
| generate local style edits with diffusion pipeline instead of AdaIN Stylized COCO (object) | Local style benchmark · Local object style | CLIP Acc: 0.574 -> 0.906 | Improved |
| generate local style edits with diffusion pipeline instead of AdaIN Stylized COCO (object) | Local style benchmark · Local object style | DINO Dist: 0.075 -> 0.002 | Improved |
| generate local background style edits with diffusion pipeline instead of AdaIN | Local style benchmark · Local background style | CLIP Acc: 0.678 -> 0.975 | Improved |
| increase object size editing strength from 0.2 to 0.4 | PSPNet · Pascal-EA | mIoU: 64.97 -> 62.83 | Hurt |
| increase object position editing strength from 0.2 to 0.4 | PSPNet · Pascal-EA | mIoU: 64.98 -> 63.67 | Hurt |
Closest papers by meaning
- Benchmarking Segmentation Models with Mask-Preserved Attribute Editing (CVPR 2024 · similarity 0.90)
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.87)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.87)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.87)
- MaSS13K: A Matting-level Semantic Segmentation Benchmark (CVPR 2025 · similarity 0.87)
- Scaling up Image Segmentation across Data and Tasks (CVPR 2025 · similarity 0.87)
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.
