Research map / Detection and segmentation / CNN features
Clustered Object Detection in Aerial Images
Fan Yang, Heng Fan, Peng Chu, Erik Blasch, Haibin Ling
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 cluster proposal network (CPNet) instead of even image partition (ablation) | FRCNN+FPN ResNeXt101 · VisDrone val | AP: 24.4 -> 27.6 | Improved |
| add CPNet instead of even image partition (fewer chips) (ablation) | FRCNN+FPN+EIP · VisDrone val | images forwarded: 3,288 -> 1,945 | Improved |
| add ScaleNet to CPNet and DetecNet (ablation) | DetecNet+CPNet ResNet50 · VisDrone val | AP: 25.6 -> 26.7 | Improved |
| add ScaleNet to CPNet and DetecNet (ablation) | DetecNet+CPNet ResNet101 · VisDrone val | AP: 25.3 -> 26.7 | Improved |
| add ScaleNet to CPNet and DetecNet (ablation) | DetecNet+CPNet ResNeXt101 · VisDrone val | AP_s: 17.5 -> 19.1 | Improved |
| add ScaleNet (partition and padding) raises processed chips (ablation) | DetecNet+CPNet ResNet50 · VisDrone val | images forwarded: 1,945 -> 2,716 | Hurt |
| add even image partition (EIP) at test time | FRCNN+FPN ResNet50 · VisDrone val | AP_l: 54.7 -> 30.0 | Hurt |
| add even image partition (EIP) at test time | FRCNN+FPN ResNeXt101 · VisDrone val | AP: 21.8 -> 24.4 | Improved |
| add even image partition (EIP) at test time | FRCNN+FPN ResNet50 · UAVDT | AP: 11.0 -> 6.6 | Hurt |
| replace EIP with cluster-based crops (ClusDet) | FRCNN+FPN ResNet50 · UAVDT | AP: 6.6 -> 13.7 | Improved |
| raise TopN above 4 (ablation) | ClusDet · VisDrone val | AP: fluctuates around AP=27 | No clear effect |
| add multi-scale inference and box voting | ClusDet ResNeXt101 · VisDrone val | AP: 28.4 -> 32.4 | Improved |
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.
- R-FCN — R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016) · better on 1 of 1 shared result
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
- Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges (TPAMI 2021 · similarity 0.91)
- Towards Large-Scale Small Object Detection: Survey and Benchmarks (TPAMI 2023 · similarity 0.89)
- SOOD++: Leveraging Unlabeled Data to Boost Oriented Object Detection (TPAMI 2025 · similarity 0.88)
- DOTA: A Large-Scale Dataset for Object Detection in Aerial Images (CVPR 2018 · similarity 0.87)
- RemDet: Rethinking Efficient Model Design for UAV Object Detection (AAAI 2025 · similarity 0.85)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 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.
