The codes build rotated cascade retinanet for arbitrary-oriented object detection. It supports the following datasets: DOTA, HRSC2016, ICDAR2013, ICDAR2015, UCAS-AOD, NWPU VHR-10, VOC2007, MSRA-TD500.
cd $ROOT/utils
sh make.sh
cd $ROOT/datasets/DOTA_devkit
sudo apt-get install swig
swig -c++ -python polyiou.i
python setup.py build_ext --inplace
pip install requirements.txt
pip install git+git://github.com/lehduong/torch-warmup-lr.git
python demo.py
- prepare dataset and move it into the
$ROOTdirectory. - generate imageset files:
cd $ROOT/datasets
python generate_imageset.py
- start training:
python train.py
prepare labels, take hrsc for example:
cd $ROOT/datasets/evaluate
python hrsc2gt.py
start evaluation:
python eval.py
