Research map / Detection and segmentation / Open-set and few-shot detection
Auto-Vocabulary Semantic Segmentation
Osman Ülger, Maksymilian Kulicki, Yuki Asano, Martin R. Oswald
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 CRF and majority-filter denoising after clustering (ablation) | BBoost · PASCAL VOC | mIoU: 70.4 -> 87.1 | Improved |
| add caption filtering (noun extraction, WordNet) (ablation) | AutoSeg · PASCAL VOC | mIoU: 3.8 -> 87.1 | Improved |
| use LAVE mapper instead of Sentence-BERT (ablation) | AutoSeg evaluation · PASCAL VOC | mIoU: 73.9 -> 87.1 | Improved |
| use LAVE mapper instead of CLIP (ablation) | AutoSeg evaluation · PASCAL VOC | mIoU: 77.5 -> 87.1 | Improved |
| replace plain BLIP embeddings with BBoost clusters as X-Decoder guidance (ablation) | AutoSeg · PASCAL VOC | c-mIoU: 38.0 -> 71.8 | Improved |
| replace plain BLIP embeddings with BBoost clusters as X-Decoder guidance (ablation) | AutoSeg · Cityscapes | c-mIoU: 29.2 -> 35.8 | Improved |
| increase captioning cycles from 1 to 25 (ablation) | AutoSeg · Cityscapes | mIoU: 28.2 -> 30.0 | Improved |
| increase captioning cycles from 1 to 10 (ablation) | AutoSeg · PASCAL VOC | mIoU: 85.3 -> 87.1 | Improved |
| increase captioning cycles from 1 to 10 (ablation) | AutoSeg · ADE20K | mIoU: 6.0 -> 5.1 | Hurt |
| replace BLIP+BBoost captions with LLaVA captions for X-Decoder | AutoSeg · PASCAL VOC | mIoU: 87.1 -> 56.7 | Hurt |
| replace local embedding captioning with SAM crop captioning | AutoSeg · PASCAL VOC | mIoU: 87.1 -> 41.1 | Hurt |
| replace LAVE mapper with manual mapper | AutoSeg · PASCAL VOC | mIoU: 87.1 -> 88.2 | No clear effect |
Closest papers by meaning
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.91)
- USE: Universal Segment Embeddings for Open-Vocabulary Image Segmentation (CVPR 2024 · similarity 0.91)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
- Open-Vocabulary Semantic Segmentation with Image Embedding Balancing (CVPR 2024 · similarity 0.90)
- Distilling Spectral Graph for Object-Context Aware Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.89)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · 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.
