Disclaimer

This is the official repo of paper Learning RoI Transformer for Detecting Oriented Objects in Aerial Images

This code is based on deformable convolution network

We refactored the code and retrained the model. There are slight differences in the final accuracy.

mmdetection version is finished, it is more faster and accurate, we recommend you to use the new version.

Requirements: Software

1. MXNet from the offical repository.

2. Python 2.7. We recommend using Anaconda2 as it already includes many common packages. We do not support Python 3 yet, if you want to use Python 3 you need to modify the code to make it work.

3. Python packages might missing: cython, opencv-python >= 3.2.0, easydict. If pip is set up on your system, those packages should be able to be fetched and installed by running

   pip install -r requirements.txt
   

4. For Windows users, Visual Studio 2015 is needed to compile cython module.

Installation

1. Clone the RoI Transformer repository, and we'll call the directory that you cloned RoI Transformer as ${RoI_ROOT}

git clone https://github.com/dingjiansw101/RoITransformer_DOTA.git

2. For Windows users, run cmd .\init.bat. For Linux user, run sh ./init.sh. The scripts will build cython module automatically and create some folders.

3. Install MXNet:

   Note: The MXNet's Custom Op cannot execute parallelly using multi-gpus after this PR. We strongly suggest the user rollback to version MXNet@(commit 998378a) for training (following Section 3.2 - 3.5).

   Build from source (Since there are custom c++ operators, We need to complie the MXNet from source.)

   3.1 Clone MXNet and checkout to MXNet@(commit 998378a) by

   git clone --recursive https://github.com/dmlc/mxnet.git
   git checkout 998378a
   git submodule update
   # if it's the first time to checkout, just use: git submodule update --init --recursive
   

   3.2 Copy the c++ operators to MXNet source

   cp ${RoI_ROOT}/fpn/operator_cxx/* ${MXNET_ROOT}/src/operator/contrib
   

   3.3 Compile MXNet

   cd ${MXNET_ROOT}
   make -j $(nproc) USE_OPENCV=1 USE_BLAS=openblas USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1
   

   3.4 Install the MXNet Python binding by

   Note: If you will actively switch between different versions of MXNet, please follow 3.5 instead of 3.4

   cd python
   sudo python setup.py install
   

   3.5 For advanced users, you may put your Python packge into ./external/mxnet/$(YOUR_MXNET_PACKAGE), and modify MXNET_VERSION in ./experiments/rfcn/cfgs/*.yaml to $(YOUR_MXNET_PACKAGE). Thus you can switch among different versions of MXNet quickly.

4. Compile dota_kit

   sudo apt-get install swig
   cd ${RoI_ROOT}/dota_kit
   swig -c++ -python polyiou.i
   python setup.py build_ext --inplace
   cd ${RoI_ROOT}/dota_kit/poly_nms_gpu
   make -j16
   

Prepare DOTA Data:

1. Prepare script put your original dota data (before split) in path_to_data make sure it looks like

   path_to_data/train/images,
   path_to_data/train/labelTxt,
   path_to_data/val/images,
   path_to_data/val/labelTxt,
   path_to_data/test/images
   
   cd ${RoI_ROOT}/prepare_data
   python prepare_data.py --data_path path_to_data --num_process 32
   

2. Create soft link

   cd ${RoI_ROOT}
   mkdir -p data
   cd data
   ln -s path_to_data dota_1024
   

Pretrained Models

We provide trained convnet models.

1. To use the demo with our pre-trained RoI Transformer models for DOTA, please download manually from Google Drive, or BaiduYun (Extraction code: fucc) and put it under the following folder. Make sure it look like this:

       ./output/rcnn/DOTA/resnet_v1_101_dota_RoITransformer_trainval_rcnn_end2end/train/rcnn_dota-0040.params
       ./output/fpn/DOTA/resnet_v1_101_dota_rotbox_light_head_RoITransformer_trainval_fpn_end2end/train/fpn_DOTA_oriented-0008.params
   

Training & Testing

cd ${RoI_ROOT}

1. training Please download ImageNet-pretrained ResNet-v1-101 model manually from OneDrive, or BaiduYun, or Google drive, and put it under folder ./model. Make sure it look like this:

   ./model/pretrained_model/resnet_v1_101-0000.params
   

   Start training (we use the Light-head R-CNN + RoI Transformer (without FPN) for example, you may choose other models)

     sh train_dota_light_RoITransformer.sh
   

2. testing

   Start testing

   sh test_dota_light_RoITransformer.sh
   

---

© Microsoft, 2017. Licensed under an MIT license.

If you find RoI Transformer and DOTA data useful in your research, please consider citing:

@inproceedings{ding2019learning,
  title={Learning RoI Transformer for Oriented Object Detection in Aerial Images},
  author={Ding, Jian and Xue, Nan and Long, Yang and Xia, Gui-Song and Lu, Qikai},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={2849--2858},
  year={2019}
}
@inproceedings{xia2018dota,
  title={DOTA: A large-scale dataset for object detection in aerial images},
  author={Xia, Gui-Song and Bai, Xiang and Ding, Jian and Zhu, Zhen and Belongie, Serge and Luo, Jiebo and Datcu, Mihai and Pelillo, Marcello and Zhang, Liangpei},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={3974--3983},
  year={2018}
}