Oriented detection in aerial and satellite images

oriented-det is an open-source PyTorch framework for oriented and rotated object detection in aerial and satellite images.

In Earth Observation, objects often have a clear heading: ships in a harbour, aircraft on an apron, vehicles in a parking lot. Axis-aligned boxes include neighbours and background. Rotated (oriented) boxes follow the object footprint instead.

Links Link to heading

• GitHub — source, configs, and pretrained zoo
• Documentation — install, training, inference, ONNX export
• PyPI — pip install oriented-det
• Optical satellite demo — three detectors on real scenes
• Tutorial series — releases, metrics, and hands-on write-ups