Research map / Detection and segmentation / CNN features
SPAR: Single-Pass Any-Resolution ViT for Open-vocabulary Segmentation
Naomi Kombol, Ivan Martinović, Siniša Šegvić, Giorgos Tolias
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 |
|---|---|---|---|
| distill sliding-window teacher into single-pass student (SPAR) (ablation) | SigLIP2 ViT-B-16 · SigLIP2 OVS Mean6 | mIoU: 33.1 -> 43.6 | Improved |
| distill into single-pass student (SPAR) (ablation) | OpenCLIP ViT-B-16 · SigLIP2 OVS Mean6 | mIoU: 27.7 -> 34.4 | Improved |
| distill into single-pass student (SPAR) (ablation) | DINOv3.txt ViT-L-16 · SigLIP2 OVS Mean6 | mIoU: 43.8 -> 44.4 | Improved |
| SPAR student vs sliding-window teacher | SigLIP2 ViT-B-16 · SigLIP2 OVS Mean6 | mIoU: 41.2 -> 43.6 | Improved |
| add LPOSS label propagation on SPAR (ablation) | SPAR SigLIP2 · SigLIP2 OVS Mean6 | mIoU: 43.6 -> 46.7 | Improved |
| add AnyUp upsampling to SPAR (ablation) | SPAR SigLIP2 · SigLIP2 OVS Mean6 | mIoU: 43.6 -> 46.8 | Improved |
| add AnyUp upsampling to SPAR (ablation) | SPAR DINOv3.txt · Mean6 | mIoU: 44.4 -> 43.8 | No clear effect |
| tune only the last 2 blocks instead of all parameters (ablation) | SPAR SigLIP2 · SigLIP2 OVS Mean6 | mIoU: 42.5 -> 43.6 | Improved |
| tune only patch projection (ablation) | SPAR SigLIP2 · SigLIP2 OVS Mean6 | mIoU: 43.6 -> 38.0 | Hurt |
| use SA-1B 25k instead of 5% subset (ablation) | SPAR SigLIP2 · SigLIP2 OVS Mean6 | mIoU: 41.1 -> 43.6 | Improved |
| double distillation data from 25k to 50k (ablation) | SPAR SigLIP2 · SigLIP2 OVS Mean6 | mIoU: 43.6 -> 43.6 | No clear effect |
| distill on ADE20k only instead of ADE20k+CS+VOC (ablation) | SPAR SigLIP2 · Cityscapes | mIoU: 39.3 -> 37.4 | Hurt |
Closest papers by meaning
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.87)
- RegionGPT: Towards Region Understanding Vision Language Model (CVPR 2024 · similarity 0.86)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
- A Simple Framework for Open-Vocabulary Zero-Shot Segmentation (ICLR 2025 · similarity 0.86)
- Your ViT is Secretly an Image Segmentation Model (CVPR 2025 · similarity 0.86)
- SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentation (CVPR 2024 · 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.
