Disclaimer

This is the official repo of paper _DOTA: A Large-scale Dataset for Object Detection in Aerial Images_. This repo contains code for training Faster R-CNN on oriented bounding boxes and horizontal bounding boxes as reported in our paper.

This code is mostly modified by Zhen Zhu and Jian Ding.

If you use these code in your project, please contain this repo in your paper or license. Please also cite our paper:

DOTA: A Large-scale Dataset for Object Detection in Aerial Images
Gui-Song Xia*, Xiang Bai*, Jian Ding, Zhen Zhu, Serge Belongie, Jiebo Luo, Mihai Datcu, Marcello Pelillo, Liangpei Zhang
In CVPR 2018. (* equal contributions)

The code is built upon a fork of Deformble Convolutional Networks. We use the Faster-RCNN part of it and make some modifications based on Faster-RCNN to regress a quadrangle. More details can be seen in our paper.

Requirements: Software

1. MXNet from the offical repository. We tested our code on MXNet@(commit 62ecb60). Due to the rapid development of MXNet, it is recommended to checkout this version if you encounter any issues.

2. Python 2.7. We recommend using Anaconda2 to manage the environments and packages.

3. Some python packages: 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 Cython
pip install opencv-python==3.2.0.6
pip install easydict==1.6

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

Requirements: Hardware

Any NVIDIA GPUs with at least 4GB memory should be sufficient.

Installation

1. Clone the repository ~~~ git clone https://github.com/jessemelpolio/Faster_RCNN_for_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.

Demo & Deformable Model

We provide trained convnet models, including Faster R-CNN models trained on DOTA.

1. To use the demo with our pre-trained faster-rcnn models for DOTA, please download manually from Google Drive, or BaiduYun and put it under the following folder.

   Make sure it look like this:

   ./output/rcnn/DOTA_quadrangle/DOTA_quadrangle/train/rcnn_DOTA_quadrangle-0059.params
   ./output/rcnn/DOTA/DOTA/train/rcnn_DOTA_aligned-0032.params

Preparation for Training & Testing

1. Please download DOTA dataset, use the DOTA_devkit to split the data into patches. And make sure the split images look like this:

./path-to-dota-split/images
./path-to-dota-split/labelTxt
./path-to-dota-split/test.txt
./path-to-dota-split/train.txt

The test.txt and train.txt are name of the subimages(without suffix) for train and test respectively.

1. 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

Usage

1. All of our experiment settings (GPU #, dataset, etc.) are kept in yaml config files at folder ./experiments/faster_rcnn/cfgs.

2. Set the "dataset_path" and "root_path" in DOTA.yaml and DOTA_quadrangle.yaml. The "dataset_path" should be the father folder of "images" and "labelTxt". The "root_path" is the path you want to save the cache data.

3. Set the scales and aspect ratios as your wish in DOTA.yaml and DOTA_quadrangle.yaml.

4. To conduct experiments, run the python scripts with the corresponding config file as input. For example, train and test on quadrangle in an end-to-end manner, run

   python experiments/faster_rcnn/rcnn_dota_quadrangle_e2e.py --cfg experiments/faster_rcnn/cfgs/DOTA_quadrangle.yaml

5. Please find more details in config files and in our code.

Misc.

Code has been tested under: - Ubuntu 14.04 with 4 Pascal Titan X GPUs and 32 Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz

Cite

If you use our project, please cite:

@article{xia2017dota,
  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},
  journal={arXiv preprint arXiv:1711.10398},
  year={2017}
}