Research map / Detection and segmentation / Open-set and few-shot detection
Detecting Everything in the Open World: Towards Universal Object Detection
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 |
|---|---|---|---|
| Train the shared backbone with COCO and O365 using partitioned label spaces instead of a unified label space (ablation) | UniDetector · LVIS v0.5 | AP: 20.9 -> 22.2 | Helped |
| Add OpenImages to partitioned COCO and O365 training (ablation) | UniDetector · LVIS v0.5 | AP: 22.2 -> 23.5 | Helped |
| Add Mosaic to unified COCO and O365 training (ablation) | UniDetector · LVIS v0.5 | AP: 20.9 -> 21.4 | Helped |
| Add pseudo labels to unified COCO and O365 training (ablation) | UniDetector · LVIS v0.5 | AP: 20.9 -> 20.8 | No clear effect |
| Decouple proposal generation and RoI classification (ablation) | UniDetector · LVIS v0.5 | AP: 18.1 -> 19.1 | Helped |
| Use a class-agnostic Faster RCNN, including RPN and RoI head, rather than a single RPN for proposals (ablation) | Decoupled UniDetector · LVIS v0.5 | AP: 19.1 -> 19.7 | Helped |
| Use OLN rather than a single RPN for proposals (ablation) | Decoupled UniDetector · LVIS v0.5 | AP: 19.1 -> 19.7 | Helped |
| Use CLN combining class-agnostic classification and localization confidence for proposals (ablation) | Decoupled UniDetector · LVIS v0.5 | AP: 19.7 -> 21.2 | Helped |
| Apply probability calibration using category prior probabilities (ablation) | UniDetector · LVIS v0.5 novel classes | box AP and mask AP: more than 2% | Helped |
| Apply probability calibration using category prior probabilities (ablation) | UniDetector · LVIS v0.5 base classes | AP | No clear effect |
| Train on heterogeneous datasets rather than COCO alone | UniDetector · LVIS v0.5 | AP: 16.4 -> 22.2 | Helped |
Closest papers by meaning
- OW-OVD: Unified Open World and Open Vocabulary Object Detection (CVPR 2025 · similarity 0.89)
- Open-World Objectness Modeling Unifies Novel Object Detection (CVPR 2025 · similarity 0.89)
- Exploring Orthogonality in Open World Object Detection (CVPR 2024 · similarity 0.88)
- Detecting Open World Objects via Partial Attribute Assignment (CVPR 2025 · similarity 0.88)
- The All-Seeing Project: Towards Panoptic Visual Recognition and Understanding of the Open World (ICLR 2024 · similarity 0.87)
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 2024 · 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.