Vision Meets Drones: A Challenge
Vision Meets Drones: A Challenge

The VisDrone dataset consists of 400 video clips formed by 265,228 frames and 10,209 static images

Overview Download
Task: Object Tracking
Task: Object Tracking

In particular, for an input video sequence and the initial bounding box of the target object in the first frame, the challenge requires a participating algorithm to locate the target bounding boxes in the subsequent video frames.

Overview Download
Task: Object Detection
Task: Object Detection

In object detection task, we focus on ten object categories of interest including pedestrian, person, car, van, bus, truck, motor, bicycle, awning-tricycle, and tricycle. Some rarely occurring special vehicles.

Overview Download
Task: Crowd Counting
Task: Crowd Counting

The challenge will provide 112 challenging sequences, including 82 video sequences for training (2,420 frames in total), and 30 sequences for testing (900 frames in total)

Overview Download

Drones, or general UAVs, equipped with cameras have been fast deployed to a wide range of applications, including agricultural, aerial photography, fast delivery, and surveillance. Consequently, automatic understanding of visual data collected from these platforms become highly demanding, which brings computer vision to drones more and more closely. We are excited to present a large-scale benchmark with carefully annotated ground-truth for various important computer vision tasks, named VisDrone, to make vision meet drones.

The VisDrone2020 dataset is collected by the AISKYEYE team at Lab of Machine Learning and Data Mining , Tianjin University, China. The benchmark dataset consists of 400 video clips formed by 265,228 frames and 10,209 static images, captured by various drone-mounted cameras, covering a wide range of aspects including location (taken from 14 different cities separated by thousands of kilometers in China), environment (urban and country), objects (pedestrian, vehicles, bicycles, etc.), and density (sparse and crowded scenes). Note that, the dataset was collected using various drone platforms (i.e., drones with different models), in different scenarios, and under various weather and lighting conditions. These frames are manually annotated with more than 2.6 million bounding boxes or points of targets of frequent interests, such as pedestrians, cars, bicycles, and tricycles. Some important attributes including scene visibility, object class and occlusion, are also provided for better data utilization.

News

  • VisDrone 2021 will be organized in conjunction with ICCV 2021.
  • Aug. 28, 2020: Computer Vision for UAVs Workshop and Challenge will be held at 8:00 (UTC+1) on August 28.
  • July. 14, 2020: Evaluation server will be closed at 23:59 on July 15 (UTC+0 time).
  • July. 9, 2020: Paper submission system is available now. Paper submission deadline is delayed until July 15th. 
  • June. 26, 2020: Due to the impact of COVID-19, the submission deadline is delayed until July 15th. Each team will have additional 5 submission opportunities.

Citation

Pengfei Zhu, Longyin Wen, Dawei Du, Xiao Bian, Qinghua Hu, Haibin Ling. Vision Meets Drones: Past, Present and Future. arXiv preprint arXiv:2001.06303 (2020). Bibtex source | Abstract | PDF