Research map / Detection and segmentation / Open-set and few-shot detection
Mitigating Objectness Bias and Region-to-Text Misalignment for Open-Vocabulary Panoptic Segmentation
Nikolay Kormushev, Josip Šarić, Matej Kristan
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 COAT CLIP-conditioned objectness adjustment to FC-CLIP (ablation) | FC-CLIP · ADE20K | PQ: 26.8 -> 27.6 | Improved |
| add COAT to FC-CLIP (ablation) | FC-CLIP · COCO (seen vocabulary) | PQ: 54.4 -> 53.7 | Hurt |
| add OVR mask-to-text fine-tuning to FC-CLIP (ablation) | FC-CLIP · ADE20K | PQ: 26.8 -> 27.6 | Improved |
| combine COAT with OVR (ablation) | OVR · ADE20K | PQ: 27.6 -> 28.6 | Improved |
| combine COAT with OVR (ablation) | OVR · COCO (seen vocabulary) | PQ: 55.5 -> 54.6 | Hurt |
| ensemble mask-transformer and CLIP probabilities instead of CLIP alone (ablation) | OVRCOAT · ADE20K | PQ: 24.9 -> 28.6 | Improved |
| add COAT to the mask classifier (ablation) | Ensemble classifier · ADE20K | PQ: 27.6 -> 28.6 | Improved |
| OVR fine-tune CLIP instead of pretrained CLIP with oracle masks (ablation) | CLIP ConvNeXt-L · ADE20K oracle masks | PQ: 41.8 -> 46.4 | Improved |
| add representation consistency or Gram matrix loss to OVR (ablation) | OVR · ADE20K | PQ: 27.6 -> 27.7 | No clear effect |
| freeze CLIP image encoder final layers instead of fully unfreezing (ablation) | OVR · ADE20K | PQ: 27.2 -> 27.6 | Improved |
| add COAT on top of OVR for semantic segmentation (ablation) | OVR · ADE20K A-150 | mIoU: 34.3 -> 33.7 | Hurt |
| increase CLIP trust factor gamma from 0.5 to 1.0 (ablation) | COAT · ADE20K | PQ: 28.6 -> 22.8 | Hurt |
Closest papers by meaning
- Open-vocabulary Panoptic Segmentation with Embedding Modulation (ICCV 2023 · similarity 0.90)
- Open-Vocabulary Camouflaged Object Segmentation (ECCV 2024 · similarity 0.89)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.89)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.88)
- ProxyDet: Synthesizing Proxy Novel Classes via Classwise Mixup for Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.87)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 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.
