csuhan/ReDet
ReDet is a rotation-equivariant object detector for aerial imagery that reduces model parameters by explicitly encoding orientation through equivariant networks.
Not currently ranked — collecting fresh signals.
star history
ReDet implements a rotation-equivariant detector for object detection in aerial images, addressing the challenge of arbitrary object orientations. It incorporates rotation-equivariant networks to extract orientation-aware features, reducing redundant parameters compared to standard CNN approaches. The method uses rotation-invariant RoI alignment to produce consistent features regardless of object orientation. This is a research implementation built on mmdetection for evaluating aerial object detection performance.
Frequently asked
- What is csuhan/ReDet?
- ReDet is a rotation-equivariant object detector for aerial imagery that reduces model parameters by explicitly encoding orientation through equivariant networks.
- Is ReDet open source?
- Yes — csuhan/ReDet is open source, released under the Apache-2.0 license.
- What language is ReDet written in?
- csuhan/ReDet is primarily written in Python.
- How popular is ReDet?
- csuhan/ReDet has 424 stars on GitHub.
- Where can I find ReDet?
- csuhan/ReDet is on GitHub at https://github.com/csuhan/ReDet.