Research map / Detection and segmentation / Open-set and few-shot detection
Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges
Jian Ding, Nan Xue, Gui-Song Xia, Xiang Bai, Wen Tao Yang, Michael Ying Yang, Serge Belongie, Jiebo Luo, Mihai P. Datcu, Marcello Pelillo, Liangpei Zhang
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 high patch overlap (512 instead of 200) (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v1.5 | OBB mAP: 65.03 -> 67.57 | Improved |
| add multi-scale training (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v1.5 | OBB mAP: 67.57 -> 69.44 | Improved |
| add multi-scale testing (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v1.5 | OBB mAP: 69.44 -> 73.62 | Improved |
| add rotation training (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v1.5 | OBB mAP: 73.62 -> 76.43 | Improved |
| add rotation testing (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v1.5 | OBB mAP: 76.43 -> 77.60 | Improved |
| replace RoI Align with RoI Transformer (ablation) | Faster R-CNN OBB · DOTA-v2.0 test-dev | OBB mAP: 47.31 -> 52.81 | Improved |
| replace RoI Align with deformable RoI pooling (Dpool) (ablation) | Faster R-CNN OBB · DOTA-v2.0 test-dev | OBB mAP: 47.31 -> 48.77 | Improved |
| use mask head instead of OBB head (Mask R-CNN vs Faster R-CNN H-OBB) | Faster R-CNN · DOTA-v2.0 test-dev | OBB mAP: +0.57 points; 4 fps slower | Improved |
| increase proposals from 1,000 to 10,000 (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v2.0 test-dev | OBB mAP: 51.72 -> 53.92; speed 14.4 -> 5.7 fps | Improved |
| increase proposals from 1,000 to 10,000 (ablation) | Faster R-CNN OBB · DOTA-v2.0 test-dev | OBB mAP: 47.10 -> 48.49; speed 15.8 -> 7.8 fps | Improved |
| exclude small instances (area <= 50 and max(w,h) <= 10) (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v2.0 | HBB mAP: 51.08 / 51.35 / 50.71 across thresholds | No clear effect |
| filter more small instances (area <= 80, max(w,h) <= 12 vs <= 10) (ablation) | Faster R-CNN OBB + RoI Transformer · DOTA-v2.0 | HBB mAP: 51.35 -> 50.71 | Hurt |
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 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
- DOTA: A Large-Scale Dataset for Object Detection in Aerial Images (CVPR 2018 · similarity 0.95)
- Towards Large-Scale Small Object Detection: Survey and Benchmarks (TPAMI 2023 · similarity 0.91)
- Clustered Object Detection in Aerial Images (ICCV 2019 · similarity 0.91)
- Multiview Aerial Visual RECognition (MAVREC): Can Multi-view Improve Aerial Visual Perception? (CVPR 2024 · similarity 0.88)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.87)
- Zero-Shot Aerial Object Detection with Visual Description Regularization (AAAI 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.
