This is the official website of the VISO (VIdeo Satellite Objects) dataset. [Google Drive][BaiduYun](Sharing code: viso)

This dataset is a large-scale dataset for moving object detection and tracking in satellite videos, which consists of 40 satellite videos captured by Jilin-1 satellite platforms. Each image has a resolution of 12000x5000 and contains a great number of objects with different scales. Four common types of vechicles, including plane, car, ship, and train, are manually-labeled. A total of 853,911 instances are labeled by axis-aligned bounding boxes.

We also build a new satellite video benchmark to fairly and extensively evaluate the performance of existing methods in several sub-tasks, including moving object detection, single-object tracking, and multi-object tracking.

• Moving Object Detection:

• Single Object Tracking:

• Multiple Object Tracking:

The provided dataset has been authorized by Changguang Satellite Technology Co., Ltd.. Licensed under the CC BY-NC-SA 4.0 license, see LICENSE.

If you find our work useful in your research, please consider citing:

    @article{yin2021detecting,
      title={Detecting and Tracking Small and Dense Moving Objects in Satellite Videos: A Benchmark},
      author={Yin, Qian and Hu, Qingyong and Liu, Hao and Zhang, Feng and Wang, Yingqian and Lin, Zaiping and An, Wei and Guo, Yulan},
      journal={IEEE Transactions on Geoscience and Remote Sensing},
      year={2021},
      publisher={IEEE}
    }

Please contact qingyong.hu@cs.ox.ac.uk if you have any questions.

1. SoTA-Point-Cloud: Deep Learning for 3D Point Clouds: A Survey
2. SensatUrban: Learning Semantics from Urban-Scale Photogrammetric Point Clouds
3. 3D-BoNet: Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds
4. SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration
5. SQN: Weakly-Supervised Semantic Segmentation of Large-Scale 3D Point Clouds with 1000x Fewer Labels