Rotation Dense Feature Pyramid Networks

Recommend improved code： https://github.com/DetectionTeamUCAS

A Tensorflow implementation of R-DFPN detection framework based on FPN.
Other rotation detection method reference R2CNN, RRPN and R2CNN_HEAD
If useful to you, please star to support my work. Thanks.

Citation

Some relevant achievements based on this code.

@article{[yang2018position](https://ieeexplore.ieee.org/document/8464244),
	title={Position Detection and Direction Prediction for Arbitrary-Oriented Ships via Multitask Rotation Region Convolutional Neural Network},
	author={Yang, Xue and Sun, Hao and Sun, Xian and  Yan, Menglong and Guo, Zhi and Fu, Kun},
	journal={IEEE Access},
	volume={6},
	pages={50839-50849},
	year={2018},
	publisher={IEEE}
}

@article{[yang2018r-dfpn](http://www.mdpi.com/2072-4292/10/1/132),
	title={Automatic ship detection in remote sensing images from google earth of complex scenes based on multiscale rotation dense feature pyramid networks},
	author={Yang, Xue and Sun, Hao and Fu, Kun and Yang, Jirui and Sun, Xian and Yan, Menglong and Guo, Zhi},
	journal={Remote Sensing},
	volume={10},
	number={1},
	pages={132},
	year={2018},
	publisher={Multidisciplinary Digital Publishing Institute}
} 

Configuration Environment

ubuntu(Encoding problems may occur on windows) + python2 + tensorflow1.2 + cv2 + cuda8.0 + GeForce GTX 1080
If you want to use cpu, you need to modify the parameters of NMS and IOU functions use_gpu = False in cfgs.py
You can also use docker environment, command: docker pull yangxue2docker/tensorflow3_gpu_cv2_sshd:v1.0

Installation

Clone the repository

git clone https://github.com/yangxue0827/R-DFPN_FPN_Tensorflow.git    

Make tfrecord

The data is VOC format, reference here
data path format ($DFPN_ROOT/data/io/divide_data.py)

├── VOCdevkit
│   ├── VOCdevkit_train
│       ├── Annotation
│       ├── JPEGImages
│    ├── VOCdevkit_test
│       ├── Annotation
│       ├── JPEGImages

cd $R-DFPN_ROOT/data/io/    
python convert_data_to_tfrecord.py --VOC_dir='***/VOCdevkit/VOCdevkit_train/' --save_name='train' --img_format='.jpg' --dataset='ship'   

Compile

cd $R-DFPN_ROOT/libs/box_utils/
python setup.py build_ext --inplace

Demo

1、Unzip the weight $R-DFPN_ROOT/output/res101_trained_weights/*.rar
2、put images in $R-DFPN_ROOT/tools/inference_image
3、Configure parameters in $R-DFPN_ROOT/libs/configs/cfgs.py and modify the project's root directory
4、image slice

cd $R-DFPN_ROOT/tools
python inference.py    

5、big image

cd $FPN_ROOT/tools
python demo.py --src_folder=.\demo_src --des_folder=.\demo_des   

Train

1、Modify $R-DFPN_ROOT/libs/lable_name_dict/***_dict.py, corresponding to the number of categories in the configuration file
2、download pretrain weight(resnet_v1_101_2016_08_28.tar.gz or resnet_v1_50_2016_08_28.tar.gz) from here, then extract to folder $R-DFPN_ROOT/data/pretrained_weights
3、

cd $R-DFPN_ROOT/tools    
python train.py    

Test tfrecord

cd $R-DFPN_ROOT/tools     
python test.py     

eval(Not recommended, Please refer here)

cd $R-DFPN_ROOT/tools       
python ship_eval.py    

Summary

tensorboard --logdir=$R-DFPN_ROOT/output/res101_summary/     

Graph

Test results