Research map / Detection and segmentation / CNN features
What Makes for Effective Detection Proposals?
Jan Hosang, Rodrigo Benenson, Piotr Dollár, Bernt Schiele
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add bounding box regression | Fast R-CNN with Bing proposals · PASCAL VOC 2007 test | mAP: 37.3 -> 49.0 | Improved |
| add bounding box regression | Fast R-CNN with EdgeBoxes proposals · PASCAL VOC 2007 test | mAP: 55.4 -> 60.4 | Improved |
| add bounding box regression | LM-LLDA with MCG proposals · PASCAL VOC 2007 test | mAP: 32.5 -> 33.0 | Improved |
| replace dense sliding window with MCG proposals | LM-LLDA · PASCAL VOC 2007 test | mAP: 34.4 -> 33.0 | Hurt |
| retrain Fast R-CNN on Objectness proposals | Fast R-CNN · PASCAL VOC 2007 test | mAP: +9.1 | Improved |
| retrain Fast R-CNN on Bing proposals | Fast R-CNN · PASCAL VOC 2007 test | mAP: +6.3 | Improved |
| retrain Fast R-CNN on MCG proposals | Fast R-CNN · PASCAL VOC 2007 test | mAP: +1.8 | Improved |
| retrain Fast R-CNN on Superpixels proposals | Fast R-CNN · PASCAL VOC 2007 test | mAP: -2.0 | Hurt |
| use MCG instead of SelectiveSearch proposals | R-CNN · PASCAL VOC 2007 test | mAP: 54.6 -> 56.5 | Improved |
| use Gaussian instead of Uniform box sampling | LM-LLDA · PASCAL VOC 2007 test | mAP: 26.6 -> 28.0 | Improved |
| use EdgeBoxes instead of Bing proposals | R-CNN · PASCAL VOC 2007 test | mAP: 36.7 -> 53.0 | Improved |
| use SelectiveSearch instead of Superpixels baseline | LM-LLDA · PASCAL VOC 2007 test | mAP: 11.3 -> 32.3 | Improved |
Closest papers by meaning
- Toward Scale-Invariance and Position-Sensitive Region Proposal Networks (ECCV 2018 · similarity 0.91)
- Should All Proposals Be Treated Equally in Object Detection? (ECCV 2022 · similarity 0.90)
- Object-Proposal Evaluation Protocol is 'Gameable' (CVPR 2016 · similarity 0.90)
- You Reap What You Sow: Using Videos to Generate High Precision Object Proposals for Weakly-Supervised Object Detection (CVPR 2019 · similarity 0.88)
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016 · similarity 0.87)
- MAD: Memory-Augmented Detection of 3D Objects (CVPR 2025 · 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.
