Research map / Detection and segmentation / Open-set and few-shot detection
OVI-MAP: Open-Vocabulary Instance-Semantic Mapping
Zilong Deng, Federico Tombari, Marc Pollefeys, Johanna Wald, Daniel Barath
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 object-centric view coverage instead of pixel-counting view selection (ablation) | OVI-MAP · Replica | average VLM queries per instance: 18.7 -> 8.6 | Improved |
| use view coverage instead of pixel counting (ablation) | OVI-MAP · Replica | AP25: 33.2 -> 34.5 | Improved |
| use view coverage instead of pixel counting (ablation) | OVI-MAP · ScanNet | AP25: 24.7 -> 23.4 | No clear effect |
| use view coverage instead of random 8 views (ablation) | OVI-MAP · Replica | mIoU: 23.8 -> 26.5 | Improved |
| use CropFormer instead of SAM2 for 2D entity segmentation (ablation) | OVI-MAP · Replica | semantic mIoU: 20.2 -> 26.9 | Improved |
| use ground-truth 2D masks instead of CropFormer (ablation) | OVI-MAP · Replica | instance AP50: 50.8 -> 65.7 | Improved |
| weight multi-view features by visible pixel count instead of plain averaging (ablation) | OVI-MAP · Replica | AP50: 19.8 -> 22.0 | Improved |
| use feature clustering instead of averaging (ablation) | OVI-MAP · Replica | mIoU: 26.5 -> 25.1 | Hurt |
| use SigLIP-large instead of CLIP ViT-L as VLM backbone (ablation) | OVI-MAP · Replica | mIoU: 15.4 -> 26.9 | Improved |
| use larger SigLIP2 so400m instead of SigLIP-large (ablation) | OVI-MAP · Replica | mIoU: 26.9 -> 26.4 | No clear effect |
| reduce frames processed to meet 30 FPS | OVO-SLAM · Replica | mIoU: 24.9 -> 21.8 | Hurt |
| reduce frames processed to meet 30 FPS | OVI-MAP · Replica | mIoU: 26.5 -> 27.0 | No clear effect |
Closest papers by meaning
- O2V-Mapping: Online Open-Vocabulary Mapping with Neural Implicit Representation (ECCV 2024 · similarity 0.90)
- LangMap: A Human-Verified Benchmark for Hierarchical Open-Vocabulary Goal Navigation (NeurIPS 2026 · similarity 0.87)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.87)
- Cross-Modal and Uncertainty-Aware Agglomeration for Open-Vocabulary 3D Scene Understanding (CVPR 2025 · similarity 0.86)
- Open-Vocabulary 3D Semantic Segmentation with Foundation Models (CVPR 2024 · similarity 0.86)
- Segment Any 3D Object with Language (ICLR 2025 · 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.
