TGRS-HRRSD-Dataset: High Resolution Remote Sensing Detection (HRRSD)

I. NOTE: JPEG files are available on BaiduCloud, GoogleDrive, and an ipv6 site bt.byr.cn

• HRRSD contains 21,761 images acquired from Google Earth and Baidu Map with the spatial resolution from 0.15-m to 1.2-m.

• There are 55,740 object instances in HRRSD.

• HRRSD contains 13 categories of RSI objects.

Moreover, this dataset is divided as several subsets, image numbers in each subset are 5401 for ‘train’, 5417 for ‘val’, and 10943 for ‘test’. And ‘train-val’ subset is a merge of ‘train’ and ‘val’.

II. Mean and Std

In most current object detection systems, means and std values of datasets are required.

You may refer to:

mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375]

Moreover, you can compute the values by yourself with file OPT2017/avr_std_detection_sets.py:

$ cd your_HRRSD_path/OPT2017
$ python avr_std_detection_sets.py 500

III. Folders

Labels

• /OPT2017/Annotations: *.xml
• /OPT2017/labels: *.txt with the form of (class x y width height)

Images

• /OPT2017/JPEGImages: *.jpg

Dataset Division

• /OPT2017/ImageSets/Main: Division of the dataset.

IV. Statistics and Benchmark

Statistics

Label	Name	N_Train	N_Val	N_Trainval	N_Test	N_All	Mean Resized Scale /pixel	Resized Scale Std /pixel
1	ship	950	948	1898	1988	3886	167.44	110.37
2	bridge	1123	1121	2244	2326	4570	246.10	110.53
3	ground track field	859	856	1717	2017	3734	276.50	100.65
4	storage tank	1099	1092	2191	2215	4406	125.60	68.41
5	basketball court	923	920	1843	2033	3876	108.19	57.46
6	tennis court	1043	1040	2083	2212	4295	102.71	38.80
7	airplane	1226	1222	2448	2451	4899	113.21	67.98
8	baseball diamond	1007	1004	2011	2022	4033	231.61	117.85
9	harbor	967	964	1931	1953	3884	163.96	94.16
10	vehicle	1188	1186	2374	2382	4756	41.96	9.99
11	crossroad	903	901	1804	2219	4023	220.54	59.24
12	T junction	1066	1065	2131	2289	4420	198.71	54.88
13	parking lot	1241	1237	2478	2480	4958	122.85	54.45

In this table, N_* refers to numbers of objects. 'Train', 'Val', 'Test' are three subsets of the dataset. 'Mean Resized Scale' shows average scale of each category. 'Resized Scale Std' is the standard deviation of category scale.

Benchmark

Category	YOLO-v2 /%	Fast R-CNN /%	Fast_R-CNN_r50 + GACL-Net /%	Faster R-CNN /%	Faster_R-CNN_r50 + GACL-Net /%
Airplane	84.6	83.3	85.1	90.8	90.8
Baseball Diamond	62.2	83.6	82.6	86.9	87.2
Basketball Court	41.3	36.7	42.1	47.9	49.7
Bridge	79.0	75.1	76.7	85.5	85.6
Crossroad	43.4	67.1	68.7	88.6	88.2
Ground Track Field	94.4	90.0	89.6	90.6	90.7
Harbor	74.4	76.0	78.4	89.4	89.7
Parking Lot	45.8	37.5	39.5	63.3	65.3
Ship	78.5	75.0	74.3	88.5	88.5
Storage Tank	72.4	79.8	80.4	88.7	89.2
T Junction	46.8	39.2	38.8	75.1	75.0
Tennis Court	67.6	75.0	77.0	80.7	80.8
Vehicle	65.1	46.1	50.7	84.0	86.9
Mean AP	65.8	66.5	68.0	81.5	82.1

GACL-Net \cite{lu2020gated} is a method proposed to improve object localization performance. Title of this paper is "Gated and Axis-Concentrated Localization Network for Remote Sensing Object Detection". The suffix "r50" denotes using 50 layer Resnet.

V. FAQ

If any question is met, please contanct me with the e-mail: [email protected].

Qestion 1: AP for the "T junction" class is always NAN or 0, why?

Anwser Q1: In some object detection frameworks, there may be a piece of code like "cls_names = lower( cls_names )". This will set class names to lower case, but class names in xml files contain "T junction" where "T" is uppercase. This actually will cause several problems. The solution is using debug sofwares to find the code of changing word cases and correct it. For the dataset, I won't change the "T junction" labels in xmls currently for lacking time.

VI. Citation

If you find HRRSD dataset useful in your research, please consider citing:

@article{zhang2019hierarchical,
  title={Hierarchical and Robust Convolutional Neural Network for Very High-Resolution Remote Sensing Object Detection},
  author={Zhang, Yuanlin and Yuan, Yuan and Feng, Yachuang and Lu, Xiaoqiang},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  volume={57},
  number={8},
  pages={5535--5548},
  year={2019},
  publisher={IEEE}
}

For more comparative experimental results, please refer to:

@article{lu2020gated,
  title={Gated and Axis-Concentrated Localization Network for Remote Sensing Object Detection},
  author={Lu, Xiaoqiang and Zhang, Yuanlin and Yuan, Yuan and Feng, Yachuang},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  volume={58},
  number={1},
  pages={179--192},
  year={2020},
  publisher={IEEE}
}