Oriented Object Detection in Aerial Imagery
Detecting ships, planes, vehicles and tanks in overhead images with rotated boxes that follow each object's heading.
Tier Grades
Tier A: Frontier AI labs
stretch
Frontier labs are not observed; the actors are universities and OpenMMLab.
Tier B: Indonesian AI labs
feasible
A ResNet50 detector trained on public DOTA data fits a modest budget.
Tier C: Institution units such as ITB
feasible
A ResNet50 detector trained on public DOTA data fits a university lab.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
feasible
Rotated-box detection of vessels or tanks fits a satellite monitoring proof of concept.
Where this comes from
Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.
Anchors
- ComputeOriented R-CNN: ResNet50 backbone, 15.1 FPS at 1024x1024 on a single RTX 2080Ti; training compute not stated in the abstract.
- DataOriented R-CNN: DOTA (75.87% mAP) and HRSC2016 (96.50% mAP) benchmarks.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 5 authors on the Oriented R-CNN paper (arXiv 2108.05699).
- Time horizonMonths.
- Shelf-life signalsOpen toolkits such as MMRotate collect many detectors, so single methods are overtaken quickly.
- Pipeline roleUniversities publish detectors; Tier B and Tier C retrain them on Indonesian imagery; Tier D applies them in monitoring proofs of concept.
Where this comes from
Team size is a proxy (author counts of exemplar papers), not a headcount.
Who and What
Exemplars
Observed actors
Linked models
Links
R&D TopicOriented Object Detection in Aerial Imagery
Models
Oriented R-CNNNorthwestern Polytechnical University, geospatial
RoI TransformerWuhan University, geospatial
Oriented object detectors for aerial images (MMRotate)OpenMMLab, geospatialReDetWuhan University, geospatial+3 more
Oriented R-CNNNorthwestern Polytechnical University, geospatialRoI TransformerWuhan University, geospatialOriented object detectors for aerial images (MMRotate)OpenMMLab, geospatialReDetWuhan University, geospatial+3 moreWho Runs Itnone recorded
Models 7
Oriented object detectors for aerial images (MMRotate)OpenMMLab, geospatialproduced byOriented R-CNNNorthwestern Polytechnical University, geospatialexemplar
R3DetShanghai Jiao Tong University, geospatialproduced byReDetWuhan University, geospatialproduced byRoI TransformerWuhan University, geospatialexemplarS2A-NetWuhan University, geospatialproduced byGliding Vertexgeospatialproduced byPapers
- DOTA: A Large-scale Dataset for Object Detection in Aerial ImagesGui-Song Xia et al., IEEE Conference on Computer Vision and Pattern Recognition, 2018, 2017 founding
- Large Selective Kernel Network for Remote Sensing Object DetectionYuxuan Li et al., arXiv preprint, 2023 state of the art
- Oriented R-CNN for Object DetectionXingxing Xie et al., ICCV 2021 (per arXiv comment), 2021 state of the art
- ReDet: A Rotation-equivariant Detector for Aerial Object DetectionJiaming Han et al., CVPR2021 (per arXiv comment), 2021 state of the art
- Align Deep Features for Oriented Object DetectionJiaming Han et al., arXiv preprint, 2020 state of the art
Where this comes from
The papers to read first: where the direction began, its current best, and a survey when one exists.
Sources
- arxiv.org/abs/2108.05699 https://arxiv.org/abs/2108.05699
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Oriented Object Detection in Aerial Imagery", https://aurion.semai.id/topics/oriented-object-detection-in-aerial-imagery/, records of October 2026.