Research map / Detection and segmentation / Open-set and few-shot detection
R-FCN-3000 at 30fps: Decoupling Detection and Classification
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 |
|---|---|---|---|
| Decouple objectness detection from fine-grained classification and multiply their scores | R-FCN-3000 · ImageNet detection, 3000 classes | mAP: 34.9% vs. 19.9% | Helped |
| Use bounding-box supervision from ImageNet classification data | R-FCN-3000 · ImageNet detection, 194 classes | mAP: 43.3% vs. 34.3 | Helped |
| Use one super-class so position-sensitive filters predict generic objectness (ablation) | R-FCN-3000 · ImageNet detection, 1000 classes | mAP: 36 -> 37.3 | Hurt |
| Increase super-classes from 1 to 100 (ablation) | R-FCN-3000 · ImageNet detection, 1000 classes | inference time: 33 -> 51 ms | Hurt |
| Decouple objectness detection and RoI classification (ablation) | D-R-FCN · PASCAL VOC 2007 test | mAP@0.5: 79.5 -> 77.6 | Hurt |
| Decouple objectness detection and RoI classification (ablation) | D-R-FCN · PASCAL VOC 2007 test | mAP@0.7: 65.3 -> 63.8 | Hurt |
| Exclude held-out classes from objectness training (ablation) | Generalized Objectness · ImageNet detection, held-out 20 classes | mAP: 34.9 -> 32 | Hurt |
| Increase training classes for generalized objectness (ablation) | Generalized Objectness · ImageNet detection, held-out 20 classes | mAP: The initial gap of 9.7% ... drops to 0.3% | Helped |
| Use RPN scores instead of R-FCN objectness (ablation) | RPN objectness baseline · PASCAL VOC 2007 test | mAP@0.7: 12.7 | Hurt |
| Add bounding-box regression to RPN objectness (ablation) | RPN objectness baseline · PASCAL VOC 2007 test | mAP@0.7: 12.7 -> 47.8 | Helped |
| Change RPN positive assignment overlap from 0.7 to 0.5 (ablation) | RPN objectness baseline · PASCAL VOC 2007 test | mAP@0.5: 47.3 -> 52.0 | Helped |
| Use C5 rather than C4 features for RPN objectness (ablation) | RPN objectness baseline · PASCAL VOC 2007 test | mAP@0.5: 66.8 -> 70.6 | Helped |
Closest papers by meaning
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.90)
- R-FCN++: Towards Accurate Region-Based Fully Convolutional Networks for Object Detection (AAAI 2018 · similarity 0.89)
- Open-World Objectness Modeling Unifies Novel Object Detection (CVPR 2025 · similarity 0.85)
- UN-DETR: Promoting Objectness Learning via Joint Supervision for Unknown Object Detection (AAAI 2025 · similarity 0.85)
- Exploring Orthogonality in Open World Object Detection (CVPR 2024 · similarity 0.85)
- YOLOv10: Real-Time End-to-End Object Detection (NeurIPS 2024 · similarity 0.85)
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.