Skip to content
This is a tensorflow implementation of R2CNN++: Multi-Dimensional Attention Based Rotation Invariant Detector with Robust Anchor Strategy.
Branch: master
Clone or download
Latest commit 9396caf Nov 19, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
DOTA.png first commit Nov 20, 2018
README.md first commit Nov 20, 2018

README.md

R2CNN-Plus-Plus_Tensorflow (Not available now.)

Abstract

This is a tensorflow implementation of R2CNN++: Multi-Dimensional Attention Based Rotation Invariant Detector with Robust Anchor Strategy.
This project is based on R2CNN, and completed by YangXue and YangJirui.

DOTA test results

1

Comparison

Part of the results are from DOTA paper.

Task1 - Oriented Leaderboard

Approaches mAP PL BD BR GTF SV LV SH TC BC ST SBF RA HA SP HC
SSD 10.59 39.83 9.09 0.64 13.18 0.26 0.39 1.11 16.24 27.57 9.23 27.16 9.09 3.03 1.05 1.01
YOLOv2 21.39 39.57 20.29 36.58 23.42 8.85 2.09 4.82 44.34 38.35 34.65 16.02 37.62 47.23 25.5 7.45
R-FCN 26.79 37.8 38.21 3.64 37.26 6.74 2.6 5.59 22.85 46.93 66.04 33.37 47.15 10.6 25.19 17.96
FR-H 36.29 47.16 61 9.8 51.74 14.87 12.8 6.88 56.26 59.97 57.32 47.83 48.7 8.23 37.25 23.05
FR-O 52.93 79.09 69.12 17.17 63.49 34.2 37.16 36.2 89.19 69.6 58.96 49.4 52.52 46.69 44.8 46.3
R2CNN 60.67 80.94 65.75 35.34 67.44 59.92 50.91 55.81 90.67 66.92 72.39 55.06 52.23 55.14 53.35 48.22
RRPN 61.01 88.52 71.20 31.66 59.30 51.85 56.19 57.25 90.81 72.84 67.38 56.69 52.84 53.08 51.94 53.58
ICN 68.20 81.40 74.30 47.70 70.30 64.90 67.80 70.00 90.80 79.10 78.20 53.60 62.90 67.00 64.20 50.20
R2CNN++ 71.16 89.66 81.22 45.50 75.10 68.27 60.17 66.83 90.90 80.69 86.15 64.05 63.48 65.34 68.01 62.05

Task2 - Horizontal Leaderboard

Approaches mAP PL BD BR GTF SV LV SH TC BC ST SBF RA HA SP HC
SSD 10.94 44.74 11.21 6.22 6.91 2 10.24 11.34 15.59 12.56 17.94 14.73 4.55 4.55 0.53 1.01
YOLOv2 39.2 76.9 33.87 22.73 34.88 38.73 32.02 52.37 61.65 48.54 33.91 29.27 36.83 36.44 38.26 11.61
R-FCN 47.24 79.33 44.26 36.58 53.53 39.38 34.15 47.29 45.66 47.74 65.84 37.92 44.23 47.23 50.64 34.9
FR-H 60.46 80.32 77.55 32.86 68.13 53.66 52.49 50.04 90.41 75.05 59.59 57 49.81 61.69 56.46 41.85
R2CNN - - - - - - - - - - - - - - - -
FPN 72.00 88.70 75.10 52.60 59.20 69.40 78.80 84.50 90.60 81.30 82.60 52.50 62.10 76.60 66.30 60.10
ICN 72.50 90.00 77.70 53.40 73.30 73.50 65.00 78.20 90.80 79.10 84.80 57.20 62.10 73.50 70.20 58.10
R2CNN++ 75.35 90.18 81.88 55.30 73.29 72.09 77.65 78.06 90.91 82.44 86.39 64.53 63.45 75.77 78.21 60.11

Requirements

1、tensorflow >= 1.2
2、cuda8.0
3、python2.7 (anaconda2 recommend)
4、opencv(cv2)
5、tfplot

Download Model

1、please download resnet50_v1resnet101_v1 pre-trained models on Imagenet, put it to data/pretrained_weights.
2、please download mobilenet_v2 pre-trained model on Imagenet, put it to data/pretrained_weights/mobilenet.
3、please download trained model by this project, put it to output/trained_weights.

Data Prepare

1、please download DOTA
2、crop data, reference:

cd $PATH_ROOT/data/io/DOTA
python train_crop.py 
python val_crop.py

3、data format

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

Compile

cd $PATH_ROOT/libs/box_utils/
python setup.py build_ext --inplace
cd $PATH_ROOT/libs/box_utils/cython_utils
python setup.py build_ext --inplace

Demo

Select a configuration file in the folder (libs/configs/) and copy its contents into cfgs.py, then download the corresponding weights.

DOTA

python demo_rh.py --src_folder='/PATH/TO/DOTA/IMAGES_ORIGINAL/' 
                  --image_ext='.png' 
                  --des_folder='/PATH/TO/SAVE/RESULTS/' 
                  --save_res=False
                  --gpu='0'

Eval

python eval.py --img_dir='/PATH/TO/DOTA/IMAGES/' 
               --image_ext='.png' 
               --test_annotation_path='/PATH/TO/TEST/ANNOTATION/'
               --gpu='0'

Inference

python inference.py --data_dir='/PATH/TO/DOTA/IMAGES_CROP/'      
                    --gpu='0'

Train

1、If you want to train your own data, please note:

(1) Modify parameters (such as CLASS_NUM, DATASET_NAME, VERSION, etc.) in $PATH_ROOT/libs/configs/cfgs.py
(2) Add category information in $PATH_ROOT/libs/label_name_dict/lable_dict.py     
(3) Add data_name to line 77 of $PATH_ROOT/data/io/read_tfrecord.py 

2、make tfrecord

cd $PATH_ROOT/data/io/  
python convert_data_to_tfrecord.py --VOC_dir='/PATH/TO/VOCdevkit/VOCdevkit_train/' 
                                   --xml_dir='Annotation'
                                   --image_dir='JPEGImages'
                                   --save_name='train' 
                                   --img_format='.png' 
                                   --dataset='DOTA'

3、train

cd $PATH_ROOT/tools
python train.py

Tensorboard

cd $PATH_ROOT/output/summary
tensorboard --logdir=.

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}
} 
You can’t perform that action at this time.