open-mmlab/mmrotate

What It Solves

Rotated object detection for aerial/satellite imagery where objects have arbitrary orientations; standard horizontal bounding boxes insufficient for dense, oriented targets like vehicles, ships, aircraft.

Primary use cases

• training rotated detectors on DOTA, HRSC2016, UCAS-AOD datasets
• evaluating oriented bounding box (OBB) models
• deploying rotated detection pipelines in PyTorch

Secondary use cases

• data augmentation for oriented objects
• model conversion to ONNX/TensorRT for edge deployment
• synthetic data generation for rotated objects

When to Use

Consider when

• need oriented bounding boxes not horizontal boxes
• working with aerial/satellite datasets (DOTA, HRSC)
• require OpenMMLab ecosystem integration
• need reproducible benchmarks for rotated detection

Verify before adopting

• ? PyTorch version compatibility (check mmrotate release notes)
• ? CUDA/cuDNN version for training speed
• ? dataset license compliance for commercial use
• ? inference latency on target hardware (Jetson, x86)

Where It Fits

Stack layer perception_localization

Start Here

Adoption Checklist

• Needs verification PyTorch version compatibility (check mmrotate release notes)
• Needs verification CUDA/cuDNN version for training speed
• Needs verification dataset license compliance for commercial use
• Needs verification inference latency on target hardware (Jetson, x86)

Each check stays "needs verification" until an official source confirms it; unconfirmed items are never marked verified.

Known Limitations & Unknowns

Known limitations

• horizontal box detectors cannot be directly reused without modification
• OBB annotation tools less mature than horizontal box tools
• fewer pre-trained rotated models vs horizontal detection
• evaluation metrics ([email protected] OBB) differ from standard COCO metrics

Alternatives & Related Tools

Related tools

• mmdetection — Built on

How is it used?

Start from the recorded entry points below, then validate against the technical checklist.

• Official site https://mmrotate.readthedocs.io/en/latest/
• Repository https://github.com/open-mmlab/mmrotate
• Documentation https://mmrotate.readthedocs.io/en/latest/

Technical checklist

• OK License identified Recorded: Apache-2.0
• OK Maintenance signal Active
• OK Verification status Official confirmed
• OK Source evidence attached 1 source record(s)
• NEEDS REVIEW Latest version recorded Not recorded

Official Links

Metadata & Governance

Dataset facts

Facts above come from the official dataset card only; unconfirmed fields stay unknown.

Related Resources & Dependencies

• mmdetection — built on (verified)

Recent Activity

Related Knowledge

Collections

• Aerial Perception Models and Datasets — Source-backed aerial dataset and model repositories for detection and tracking research and prototyping.
• Edge AI Perception Starter — A starting stack for prototyping aerial perception models before edge deployment work.
• Visual Perception Models and Data — Indexed model and dataset repositories for aerial visual perception research and prototyping.