Research map / Detection and segmentation / Open-set and few-shot detection
Mask-Free OVIS: Open-Vocabulary Instance Segmentation Without Manual Mask Annotations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Generate pseudo-labels with the proposed pipeline rather than using thresholded GradCAM activations (ablation) | Mask-free OVIS · MS-COCO constrained novel | mAP50 (detection and instance segmentation): 8.6 -> 17.3; 5.2 -> 14.8 | Helped |
| Train Mask R-CNN on pseudo-labels for novel categories (ablation) | Mask-free OVIS · MS-COCO constrained novel | mAP50 (detection and instance segmentation): 17.3 -> 31.1; 14.8 -> 27.0 | Helped |
| Add pseudo-labels from base categories to training (ablation) | Mask-free OVIS · MS-COCO | mAP50 (detection; constrained novel): 31.1 -> 32.4 | Helped |
| Fine-tune on manual base annotations (ablation) | Mask-free OVIS · MS-COCO | mAP50 (detection; constrained novel): 32.4 -> 35.9 | Helped |
| Replace image-label-generated pseudo-captions with human-provided captions (ablation) | Mask-free OVIS · MS-COCO | mAP50 (detection and instance segmentation; constrained novel): 35.9 -> 36.1; 31.0 -> 31.5 | No clear effect |
| Use WSPN instead of Selective Search for proposal generation (ablation) | Mask-free OVIS · MS-COCO constrained novel | mAP50 (detection): 34.5 -> 35.9 | Helped |
| Increase iterative masking steps from the initial setting (ablation) | Mask-free OVIS pseudo-mask generation · Object detection and instance segmentation | performance | Mixed |
| Use Mask-free OVIS rather than PB-OVD for detection | MS-COCO object detection · MS-COCO constrained and generalized novel | mAP50: 32.3 -> 35.9; 30.7 -> 31.5 | Helped |
| Use Mask-free OVIS rather than XPM for instance segmentation | MS-COCO and Open Images · Constrained and generalized novel splits | mAP50: 24.0 -> 27.4; 21.6 -> 25.0; 31.6 -> 35.9; 22.7 -> 25.8 | Helped |
Closest papers by meaning
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.91)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.90)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · similarity 0.89)
- Transferable and Principled Efficiency for Open-Vocabulary Segmentation (CVPR 2024 · similarity 0.89)
- CLIP as RNN: Segment Countless Visual Concepts without Training Endeavor (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.