Research map / Detection and segmentation / Weakly supervised detection and segmentation
Box-Level Active 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 |
|---|---|---|---|
| Use ComPAS rather than the strongest listed baseline, MeanEntropy-MCDropout | VOC-sup active detection · VOC-sup, cycle 9 | mAP50: 81.40±0.35 -> 82.13±0.25 | Helped |
| Use ComPAS rather than ActiveTeacher | VOC-semi active detection · VOC-semi, cycle 4 | mAP50: 80.27±0.28 -> 81.97±0.31 | Helped |
| Use ComPAS rather than the strongest listed baseline, MeanEntropy-MCDropout | COCO-sup active detection · COCO-sup, cycle 9 | mAP: 29.37±0.21 -> 33.50±0.00 | Helped |
| Use ComPAS rather than ActiveTeacher | COCO-semi active detection · COCO-semi, cycle 9 | mAP: 33.01±0.16 -> 35.73±0.12 | Helped |
| Add an acquisition reference based on EMA chairman predictions instead of comparing committee members directly (ablation) | ComPAS acquisition on VOC-sup · VOC-sup, cycle 9 | mAP50: 81.20±0.20 -> 82.13±0.25 | Helped |
| Add the EMA chairman as the reference instead of using the detector itself (ablation) | ComPAS acquisition on VOC-sup · VOC-sup, cycle 9 | mAP50: 80.70±0.28 -> 82.13±0.25 | Helped |
| Use an EMA chairman reference rather than member-to-member assignment (ablation) | ComPAS acquisition on VOC-sup · VOC-sup acquisition | inference time per image: 0.6739 -> 0.1667 | Helped |
| Increase the input-end committee from one to four members (ablation) | sComPAS · VOC-sup, cycle 9 | mAP50: 81.70±0.26 -> 81.83±0.15 | Helped |
| Increase the input-end committee from four to ten members (ablation) | sComPAS · VOC-sup, cycle 9 | mAP50: 81.83±0.15 -> 82.13±0.25 | Helped |
| Use the classification-localization disagreement product instead of classification disagreement alone (ablation) | Box-level active acquisition · VOC-sup | mAP50 | Helped |
| Add pseudo-labels to complement sparse human annotations (ablation) | Box-level sparse training · VOC-sup | mAP50 | Helped |
| Remove pseudo-label generation from box-level sparse training (ablation) | box-Entropy-noPL · VOC-sup | mAP50 | Hurt |
Closest papers by meaning
- Plug and Play Active Learning for Object Detection (CVPR 2024 · similarity 0.90)
- Portable Active Learning for Object Detection (CVPR 2026 · similarity 0.90)
- Not All Labels Are Equal: Rationalizing the Labeling Costs for Training Object Detection (CVPR 2022 · similarity 0.88)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.87)
- Uncertainty Meets Diversity: A Comprehensive Active Learning Framework for Indoor 3D Object Detection (CVPR 2025 · similarity 0.87)
- Active Domain Adaptation with False Negative Prediction for Object Detection (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.