Research map / Detection and segmentation / CNN features
Learning to Rank Proposals for Object Detection
Zhiyu Tan, Xuecheng Nie, Qi Qian, Nan Li, Hao Li
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 |
|---|---|---|---|
| use ranking margin alpha 0.5 instead of 0 (ablation) | Rank-NMS (Faster R-CNN R50-FPN) · PASCAL VOC 2007 val (Faster R-CNN ResNet-50-FPN) | AP: 54.45 -> 58.75 | Improved |
| increase ranking margin alpha from 0.5 to 1.0 (ablation) | Rank-NMS (Faster R-CNN R50-FPN) · PASCAL VOC 2007 val (Faster R-CNN ResNet-50-FPN) | AP: 58.75 -> 57.22 | Hurt |
| use hard-pair sampling instead of full-pair (ablation) | Rank-NMS (Faster R-CNN R50-FPN) · PASCAL VOC 2007 val (Faster R-CNN ResNet-50-FPN) | AP: 56.54 -> 58.75 | Improved |
| lower positive sample ratio from 0.75 to 0.25 (ablation) | Rank-NMS (Faster R-CNN R50-FPN) · PASCAL VOC 2007 val (Faster R-CNN ResNet-50-FPN) | AP: 58.60 -> 58.75 | No clear effect |
| fuse ranking score with classification score (ablation) | NMS-Cls (Faster R-CNN R50-FPN) · MS COCO val5k | AP: 36.4 -> 38.6 | Improved |
| use ranking score alone instead of classification score (ablation) | NMS-Cls (Faster R-CNN R50-FPN) · MS COCO val5k | AP: 36.4 -> 35.4 | Hurt |
| increase fusion weight beta from 0.05 to 0.25 at alpha 0.25 (ablation) | Rank-NMS (Faster R-CNN R50-FPN) · PASCAL VOC 2007 val (Faster R-CNN ResNet-50-FPN) | AP: 57.28 -> 58.09 | Improved |
| increase fusion weight beta from 0.05 to 0.45 at alpha 0.75 (ablation) | Rank-NMS (Faster R-CNN R50-FPN) · PASCAL VOC 2007 val (Faster R-CNN ResNet-50-FPN) | AP: 58.44 -> 54.08 | Hurt |
| add Rank-NMS | Faster R-CNN R50-FPN · PASCAL VOC 2007 test | AP: 54.16 -> 58.75 | Improved |
| add Rank-NMS | Cascade R-CNN R50-FPN · PASCAL VOC 2007 test | AP: 60.09 -> 61.32 | Improved |
| add Rank-NMS | Faster R-CNN R101-FPN · PASCAL VOC 2007 test | AP: 58.22 -> 61.57 | Improved |
| add Rank-NMS | Mask R-CNN R50-FPN · MS COCO 2017 val | AP: 37.3 -> 39.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Cascade R-CNN — Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018) · better on 6 of 6 shared results
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RankDetNet: Delving Into Ranking Constraints for Object Detection (CVPR 2021 · similarity 0.91)
- Learning Non-Maximum Suppression (CVPR 2017 · similarity 0.90)
- Rank-DETR for High Quality Object Detection (NeurIPS 2023 · similarity 0.89)
- Accelerating Non-Maximum Suppression: A Graph Theory Perspective (NeurIPS 2024 · similarity 0.88)
- Soft-NMS -- Improving Object Detection With One Line of Code (ICCV 2017 · similarity 0.87)
- Bucketed Ranking-based Losses for Efficient Training of Object Detectors (ECCV 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.
