To be updated.

The codes build RetinaNet with the proposed DAL method for rotation object detection. The supported datasets include:

• DOTA
• HRSC
• ICDAR2013
• ICDAR2015
• UCAS-AOD
• NWPU VHR-10
• VOC

Note that we use only 3 horizontal perset anchors at each location on feature map for rotation detection(while 5 for IC15) . This implementation reaches 24 fps on RTX 2080 Ti.

Note that VOC07 metric is used for evaluation.

Refer to this repo.

The performance for long text detection is not good enough. The submissions are shown in the official website: Incidental Scene Text 2015.

Experiments on DOTA are implemented based on mmdetection, since the recognition ability of some classes(TC, BC, GTF) is abnormally poor. Trained models are available here.

Build the Cython and CUDA modules:

cd $ROOT/utils
sh make.sh
cd $ROOT/utils/overlaps_cuda
python setup.py build_ext --inplace

Installation for DOTA_devkit:

cd $ROOT/DOTA_devkit
sudo apt-get install swig
swig -c++ -python polyiou.i
python setup.py build_ext --inplace

You can use the following command to test a dataset. Note that weight, img_dir, dataset,hyp should be modified as appropriate.

python demo.py

1. Move the dataset to the $ROOT directory.
2. Generate imageset files for daatset division via:

cd $ROOT/datasets
python generate_imageset.py

3. Modify the configuration file hyp.py and arguments in train.py, then start training:

python train.py

Different datasets use different test methods. For UCAS-AOD/HRSC2016/VOC/NWPU VHR-10, you need to prepare labels in the appropriate format in advance. Take evaluation on HRSC2016 for example:

cd $ROOT/datasets/evaluate
python hrsc2gt.py

then you can conduct evaluation:

python eval.py

Note that :

• the script needs to be executed only once, but testing on different datasets needs to be executed again.
• the imageset file used in hrsc2gt.py is generated from generate_imageset.py.

To be updated.