We develop Wholly-WOOD (Wholly Leveraging Diversified-quality Labels for Weakly-supervised Oriented Object Detection), a weakly-supervised OOD framework, capable of wholly leveraging various labeling forms (Points, HBoxes, RBoxes, and their combination) in a unified fashion. By only using HBox for training, our Wholly-WOOD achieves performance very close to that of the RBox-trained counterpart on remote sensing and other areas, which significantly reduces the tedious efforts on labor-intensive annotation for oriented objects.
This project is the implementation of Wholly-WOOD. The code works with PyTorch 1.13+ and it is forked from MMRotate dev-1.x. MMRotate is an open-source toolbox for rotated object detection based on PyTorch. It is a part of the OpenMMLab project.
Please refer to Installation for more detailed instruction.
Please see Overview for the general introduction of MMRotate.
For detailed user guides and advanced guides, please refer to MMRotate's documentation.
The examples of training and testing Wholly-WOOD can be found here.
This repository contains the Wholly-WOOD model and our series of work on weakly-supervised OOD (i.e. H2RBox, H2RBox-v2, and Point2RBox).
Supported algorithms:
Please refer to data_preparation.md to prepare the data.
Please refer to FAQ for frequently asked questions.
This project is based on MMRotate, an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors who implement their methods or add new features, as well as users who give valuable feedbacks. We appreciate the Student Innovation Center of SJTU for providing rich computing resources at the beginning of the project. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new methods.
Coming soon