Research map / Detection and segmentation / Open-set and few-shot detection
FreeSOLO: Learning To Segment Objects Without Annotations
Check 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 |
|---|---|---|---|
| Use FreeSOLO rather than the strongest annotation-using proposal baseline, COB | Class-agnostic instance segmentation · COCO val2017 | AP50: 8.8 -> 9.8 | Helped |
| Use FreeSOLO rather than DETReg for unsupervised box discovery | Class-agnostic object detection · COCO val2017 | AP: 1.0 -> 5.5 | Helped |
| Use FreeSOLO rather than LOST* for multi-object discovery | Object discovery · COCO 20k | AP: 2.5 -> 5.6 | Helped |
| Pre-train with FreeSOLO rather than DenseCL | Instance segmentation fine-tuning with 5% fully annotated images · COCO val2017 | AP: 20.0 -> 22.0 | Helped |
| Pre-train with FreeSOLO rather than DenseCL | Instance segmentation fine-tuning with 5% segmentation masks · COCO val2017 | AP: 20.1 -> 29.9 | Helped |
| Use DenseCL instead of SimCLR for Free Mask pre-training (ablation) | Free Mask · COCO val2017 | AR: 6.1 -> 11.5 | Helped |
| Use pyramid queries instead of single-scale queries (ablation) | Free Mask · COCO val2017 | AR: 11.3 -> 11.5 | Helped |
| Perform self-training once rather than no self-training (ablation) | Self-Supervised SOLO · COCO val2017 | AP: 3.3 -> 3.7 | Helped |
| Perform a second self-training iteration (ablation) | Self-Supervised SOLO · COCO val2017 | AP: 3.7 -> 3.5 | Hurt |
| Use weakly supervised mask losses rather than full mask supervision on noisy coarse masks (ablation) | Self-Supervised SOLO · COCO val2017 | AP: 2.4 -> 3.3 | Helped |
| Add average projection loss (ablation) | Self-Supervised SOLO mask loss · COCO val2017 | AP: 2.0 -> 3.3 | Helped |
| Add max projection loss (ablation) | Self-Supervised SOLO mask loss · COCO val2017 | AP: 2.6 -> 3.3 | Helped |
Closest papers by meaning
- Exemplar-FreeSOLO: Enhancing Unsupervised Instance Segmentation With Exemplars (CVPR 2023 · similarity 0.90)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.89)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.88)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 2025 · similarity 0.87)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.86)
- Segment Anything without Supervision (NeurIPS 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.