Research map / Detection and segmentation / Open-set and few-shot detection
Language-driven Semantic Segmentation
Boyi Li, Kilian Q Weinberger, Serge Belongie, Vladlen Koltun, Rene Ranftl
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 LSeg instead of ZS3Net with ResNet101 | Zero-shot semantic segmentation · PASCAL-5i | mIoU: 38.3 -> 47.4 | Improved |
| Use LSeg instead of ZS3Net with ResNet101 | Zero-shot semantic segmentation · COCO-20i | mIoU: 21.1 -> 23.4 | Improved |
| Use LSeg with ViT-L/16 instead of HSNet with ResNet101 | Semantic segmentation · FSS-1000 | mIoU: 86.5 -> 87.8 | Improved |
| Use LSeg with ResNet101 instead of HSNet with ResNet101 | Semantic segmentation · FSS-1000 | mIoU: 86.5 -> 84.7 | Hurt |
| Add two DepthwiseBlocks rather than using none (ablation) | LSeg · ADE20K | mIoU: 37.83 -> 39.45 | Improved |
| Add two BottleneckBlocks rather than using none (ablation) | LSeg · ADE20K | mIoU: 37.83 -> 39.79 | Improved |
| Use four DepthwiseBlocks rather than two (ablation) | LSeg · ADE20K | mIoU: 39.45 -> 0.18 | Hurt |
| Use four BottleneckBlocks rather than two (ablation) | LSeg · ADE20K | mIoU: 39.79 -> 38.78 | Hurt |
| Replace CLIP ViT-B/32 text encoder with CLIP RN50x16 (ablation) | LSeg with ViT-B/32 image encoder · ADE20K | mIoU: 37.83 -> 40.36 | Improved |
| Replace CLIP ViT-B/32 text encoder with CLIP ViT-B/16 (ablation) | LSeg with ViT-B/32 image encoder · ADE20K | mIoU: 37.83 -> 38.69 | Improved |
| Replace CLIP ViT-B/32 text encoder with CLIP RN50x4 (ablation) | LSeg with ViT-B/32 image encoder · ADE20K | mIoU: 37.83 -> 38.93 | Improved |
| Use RN50x16 rather than ViT-B/32 text encoder on the fixed label set (ablation) | LSeg with ViT-L/16 image encoder · ADE20K validation | mIoU: 46.28 -> 47.25 | Improved |
Closest papers by meaning
- Exploring Simple Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.90)
- CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.90)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.89)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.89)
- Text4Seg: Reimagining Image Segmentation as Text Generation (ICLR 2025 · similarity 0.89)
- APSeg: Auto-Prompt Network for Cross-Domain Few-Shot Semantic Segmentation (CVPR 2024 · similarity 0.89)
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.
