Research map / Detection and segmentation / Weakly supervised detection and segmentation
RankSeg: Adaptive Pixel Classification with Image Category Ranking for Segmentation
Haodi He, Yuhui Yuan, Xiangyu Yue, Han Hu
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 RankSeg (multi-label head, label selection, rank-adaptive temperature) | Mask2Former Swin-L · ADE20K panoptic | PQ: 48.1 -> 48.9 | Improved |
| add multi-task multi-label classification only (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 41.85 -> 42.38 | Improved |
| add multi-task multi-label classification only (ablation) | Segmenter ViT-B · ADE20K-Full | mIoU: 17.93 -> 17.81 | No clear effect |
| add label sorting and selection to MT (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 42.38 -> 44.31 | Improved |
| add rank-adaptive temperature instead of shared temperature (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 44.31 -> 44.98 | Improved |
| add rank-adaptive temperature instead of shared temperature (ablation) | Segmenter ViT-B · VSPW | mIoU: 59.6 -> 60.1 | Improved |
| use more accurate multi-label predictor (ResNet101 to Swin-L Query2Label) (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 39.34 -> 44.42 | Improved |
| use joint multi-task instead of independent single-task scheme (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 44.42 -> 44.98 | Improved |
| increase selected label count kappa from 50 to 171 (all labels) (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 44.98 -> 44.20 | Hurt |
| increase multi-label loss weight from 10 to 20 (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 44.98 -> 44.04 | Hurt |
| use 1x transformer encoder head instead of GAP+Linear for multi-label prediction (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU: 43.19 -> 44.98 | Improved |
| derive multi-label predictions from segmentation output (mAP 58.55) (ablation) | Segmenter ViT-B · COCO-Stuff | mIoU | No clear effect |
Closest papers by meaning
- RankSEG-RMA: An Efficient Segmentation Algorithm via Reciprocal Moment Approximation (NeurIPS 2025 · similarity 0.90)
- RankSEG: A Consistent Ranking-based Framework for Segmentation (JMLR 2023 · similarity 0.89)
- Beyond Pixels: Semi-Supervised Semantic Segmentation with a Multi-scale Patch-based Multi-Label Classifier (ECCV 2024 · similarity 0.89)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.87)
- SSA-Seg: Semantic and Spatial Adaptive Pixel-level Classifier for Semantic Segmentation (NeurIPS 2024 · similarity 0.86)
- On the Relaxation of Conditional Independence Assumption for Image Segmentation (NeurIPS 2026 · 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.
