Research map / Detection and segmentation / CNN features
DeRPN: Taking a Further Step toward More General 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 |
|---|---|---|---|
| Replace RPN anchor boxes with DeRPN dimension-decomposed anchor strings | Faster R-CNN, VGG16 · PASCAL VOC 2007 test | mAP: 73.2 -> 76.5 | Helped |
| Replace RPN anchor boxes with DeRPN dimension-decomposed anchor strings | Faster R-CNN, ResNet-101 · PASCAL VOC 2012 test | mAP: 73.8 -> 76.5 | Helped |
| Use DeRPN with fixed anchor strings instead of RPN with coco-type anchors | Faster R-CNN, VGG16 · MS COCO val | AP: 24.3 -> 25.7 | Helped |
| Use DeRPN proposals instead of RPN proposals | VGG16 proposal network · PASCAL VOC 2007 test, IoU 0.7 | recall: 81.25 -> 85.51 | Helped |
| Replace RPN with DeRPN | FRCN, VGG16 · ICDAR 2013 test | F-measure (ICDAR 2013 protocol): 76.80 -> 81.86 | Helped |
| Replace RPN with DeRPN | R-FCN, ResNet-101 · ICDAR 2013 test | F-measure (DetEval): 87.34 -> 89.55 | Helped |
| Replace RPN with DeRPN and keep DeRPN anchor settings fixed | R-FCN, ResNet-101 · COCO-Text test | AP 50: 40.46 -> 53.43 | Helped |
| Use manually designed anchors instead of coco-type anchors (ablation) | RPN in R-FCN, ResNet-101 · COCO-Text test | AP 50: 40.46 -> 39.76 | Hurt |
| Replace coco-type anchors with K-means-generated anchors (ablation) | RPN in R-FCN, ResNet-101 · COCO-Text test | AP 50: 40.46 -> 42.59 | Helped |
| Replace coco-type anchors with K-means-generated anchors (ablation) | RPN in R-FCN, ResNet-101 · COCO-Text test | AP 75: 13.67 -> 13.19 | Hurt |
| Use VOC-type anchors instead of coco-type anchors (ablation) | RPN in Faster R-CNN, VGG16 · MS COCO val | AP: 24.3 -> 22.9 | Hurt |
| Use DeRPN rather than RPN for scene-text proposals | R-FCN, ResNet-101 · COCO-Text test | F-measure: 55.91 -> 57.11 | Helped |
Closest papers by meaning
- RPN Prototype Alignment for Domain Adaptive Object Detector (CVPR 2021 · similarity 0.86)
- Learning a Unified Sample Weighting Network for Object Detection (CVPR 2020 · similarity 0.86)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.86)
- Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection (NeurIPS 2020 · similarity 0.86)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.85)
- NeRF-RPN: A General Framework for Object Detection in NeRFs (CVPR 2023 · 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.