Research map / Detection and segmentation / CNN features
Libra R-CNN: Towards Balanced Learning for Object Detection
Jiangmiao Pang, Kai Chen, Jianping Shi, Huajun Feng, Wanli Ouyang, Dahua Lin
Check 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 the complete Libra R-CNN training design | Faster R-CNN with ResNet-50-FPN · COCO test-dev | AP: 36.2 -> 38.7 | Improved |
| Add the complete Libra R-CNN training design | RetinaNet with ResNet-50-FPN · COCO test-dev | AP: 35.8 -> 37.8 | Improved |
| Apply balanced training to RPN | RPN with ResNet-50-FPN · COCO val-2017 | AR^100: 42.5 -> 52.1 | Improved |
| Add IoU-balanced sampling (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 35.9 -> 36.8 | Improved |
| Add balanced feature pyramid after IoU-balanced sampling (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 36.8 -> 37.7 | Improved |
| Add balanced L1 loss after IoU-balanced sampling and balanced feature pyramid (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 37.7 -> 38.5 | Improved |
| Sample an equal number of positive examples for each ground truth (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 35.9 -> 36.1 | Improved |
| Vary IoU sampling intervals among K = 2, K = 3, and K = 5 (ablation) | IoU-balanced sampling · COCO val-2017 | AP: 36.7; 36.8; 36.7 | No clear effect |
| Use balanced feature integration without refinement (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 35.9 -> 36.3 | Improved |
| Add balanced feature pyramid to PAFPN (ablation) | PAFPN · COCO val-2017 | AP: 36.3 -> 37.2 | Improved |
| Raise localization loss weight directly to 1.5 (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 35.9 -> 36.4 | Improved |
| Raise localization loss weight directly to 2.0 (ablation) | Faster R-CNN, ResNet-50-FPN · COCO val-2017 | AP: 35.9 -> 36.2 | Hurt |
Later papers that compared with this method
5 later accepted papers report Libra R-CNN as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- PIDray: A Large-Scale X-ray Benchmark for Real-World Prohibited Item Detection (IJCV 2023 · better on 1 of 1 shared result)
- Towards Real-World Prohibited Item Detection: A Large-Scale X-Ray Benchmark (ICCV 2021 · better on 2 of 2 shared results)
- Rank & Sort Loss for Object Detection and Instance Segmentation (ICCV 2021 · better on 1 of 1 shared result)
- CircleNet for Hip Landmark Detection (AAAI 2020 · better on 1 of 1 shared result)
- FAS-Net: Construct Effective Features Adaptively for Multi-Scale Object Detection (AAAI 2020 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 4 of 4 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
- Towards Balanced Learning for Instance Recognition (IJCV 2021 · similarity 0.90)
- Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018 · similarity 0.88)
- Sparse R-CNN: An End-to-End Framework for Object Detection (TPAMI 2023 · similarity 0.88)
- Sparse R-CNN: End-to-End Object Detection With Learnable Proposals (CVPR 2021 · similarity 0.87)
- Meta R-CNN: Towards General Solver for Instance-Level Low-Shot Learning (ICCV 2019 · similarity 0.87)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · 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.
