R³Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object

Abstract

R³Det and R³Det++ are based on Focal Loss for Dense Object Detection, and it is completed by YangXue.

mmdetection version is released. We also recommend a tensorflow-based rotation detection benchmark, which is led by YangXue.

Pipeline

Latest Performance

DOTA1.0 (Task1)

Model	Backbone	Training data	Val data	mAP	Model Link	Anchor	Angle Pred.	Reg. Loss	Angle Range	lr schd	Data Augmentation	GPU	Image/GPU	Configs
RetinaNet-H	ResNet50_v1d 600->800	DOTA1.0 trainval	DOTA1.0 test	65.73	Baidu Drive (jum2)	H	Reg.	smooth L1	90	2x	×	3X GeForce RTX 2080 Ti	1	cfgs_res50_dota_v4.py
RetinaNet-H	ResNet50_v1d 600->800	DOTA1.0 trainval	DOTA1.0 test	64.17	Baidu Drive (j5l0)	H	Reg.	smooth L1	180	2x	×	3X GeForce RTX 2080 Ti	1	cfgs_res50_dota_v15.py
R3Det	ResNet50_v1d 600->800	DOTA1.0 trainval	DOTA1.0 test	70.66	Baidu Drive (30lt)	H + R	Reg.	smooth L1	90	2x	×	3X GeForce RTX 2080 Ti	1	cfgs_res50_dota_r3det_v1.py
R3Det*	ResNet101_v1d 600->800	DOTA1.0 trainval	DOTA1.0 test	73.79	Baidu Drive (d7qp)	H + R	Reg.	iou-smooth L1	90	3x	√	4X GeForce RTX 2080 Ti	1	cfgs_res101_dota_r3det_v19.py
R3Det*	ResNet152_v1d 600->800	DOTA1.0 trainval	DOTA1.0 test	74.54	Baidu Drive (73bc)	H + R	Reg.	iou-smooth L1	90	3x	√	4X GeForce RTX 2080 Ti	1	cfgs_res152_dota_r3det_v25.py
R3Det	ResNet152_v1d 600->MS (+Flip)	DOTA1.0 trainval	DOTA1.0 test	76.23 (+0.24)	model	H + R	Reg.	iou-smooth L1	90	4x	√	3X GeForce RTX 2080 Ti	1	cfgs_res152_dota_r3det_v3.py

R³Det*: R³Det with two refinement stages
Due to the improvement of the code, the performance of this repo is gradually improving, so the experimental results in other configuration files are for reference only.

Visualization

My Development Environment

docker images: docker pull yangxue2docker/yx-tf-det:tensorflow1.13.1-cuda10-gpu-py3
1、python3.5 (anaconda recommend)
2、cuda 10.0
3、opencv(cv2)
4、tfplot 0.2.0 (optional)
5、tensorflow-gpu 1.13

Download Model

Pretrain weights

1、Please download resnet50_v1, resnet101_v1, resnet152_v1, efficientnet, mobilenet_v2 pre-trained models on Imagenet, put it to data/pretrained_weights.
2、(Recommend in this repo) Or you can choose to use a better backbone (resnet_v1d), refer to gluon2TF.

• Baidu Drive, password: 5ht9.
• Google Drive

Compile

cd $PATH_ROOT/libs/box_utils/cython_utils
python setup.py build_ext --inplace (or make)

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

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/label_dict.py     
(3) Add data_name to $PATH_ROOT/data/io/read_tfrecord_multi_gpu.py  

2、Make tfrecord
For DOTA dataset:

cd $PATH_ROOT/data/io/DOTA
python data_crop.py

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

3、Multi-gpu train

cd $PATH_ROOT/tools
python multi_gpu_train_r3det.py

Test

cd $PATH_ROOT/tools
python test_dota_r3det_ms.py --test_dir='/PATH/TO/IMAGES/'  
                             --gpus=0,1,2,3,4,5,6,7  
                             -ms (multi-scale testing, optional)
                             -s (visualization, optional)

Notice: In order to set the breakpoint conveniently, the read and write mode of the file is' a+'. If the model of the same #VERSION needs to be tested again, the original test results need to be deleted.

Tensorboard

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

Citation

If this is useful for your research, please consider cite.

@article{yang2020arbitrary,
    title={Arbitrary-Oriented Object Detection with Circular Smooth Label},
    author={Yang, Xue and Yan, Junchi},
    journal={European Conference on Computer Vision (ECCV)},
    year={2020}
    organization={Springer}
}

@article{yang2019r3det,
    title={R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object},
    author={Yang, Xue and Yan, Junchi and Feng, Ziming and He, Tao},
    journal={arXiv preprint arXiv:1908.05612},
    year={2019}
}

@article{yang2020scrdet++,
    title={SCRDet++: Detecting Small, Cluttered and Rotated Objects via Instance-Level Feature Denoising and Rotation Loss Smoothing},
    author={Yang, Xue and Yan, Junchi and Yang, Xiaokang and Tang, Jin and Liao, Wenglong and He, Tao},
    journal={arXiv preprint arXiv:2004.13316},
    year={2020}
}

@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 (CVPR)},
    pages={3974--3983},
    year={2018}
}

Reference

1、https://github.com/endernewton/tf-faster-rcnn
2、https://github.com/zengarden/light_head_rcnn
3、https://github.com/tensorflow/models/tree/master/research/object_detection
4、https://github.com/fizyr/keras-retinanet