VisDrone 2019

Overview

The VisDrone2019 dataset consists of 288 video clips formed by 261,908 frames and 10,209 static images, covering a wide range of subjects: environment (urban and country), objects (pedestrian, vehicles, bicycles, etc.), and density (sparse and crowded scenes). The frames are manually annotated with more than 2.6 million bounding boxes 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.

There are 4 associated challenges with this dataset: Object Detection in Images, Object Detection in Videos, Single-Object Tracking and Multi-Object Tracking.

Associated Paper or Article

No associated paper or article has been provided.

Download

To download, one must create an account on this website before accessing the download page, where you can download several train, validationa and test dataset, in respect to the challenges proposed: Object Detection in Images, Object Detection in Videos, Single-Object Tracking or Multi-Object Tracking.

Model

No model has been provided.

Benchmarks

Benchmarks have been provided in the form of leaderboards for each of the 4 challenges. The benchmarks can be found in the "Leaderboards" section of the Evaluate drop-down menu on the dataset's official website. The evaluation criteria for each section can be found here:

• Object Detection in Images
• Object Detection in Videos
• Single-Object Tracking
• Multi-Object Tracking

Size

• Object Detection in Images: 1.44 GB Train, 0.77 GB Validation, 0.56 GB Test.
• Object Detection in Videos: 7.53 GB Train, 1.49 GB Validation, 4.84 GB Test.
• Single-Object Tracking: 20.37 GB Train, 12.56 GB Validation, 46.71 GB Test.
• Multi-Object Tracking: 7.53 GB Train, 1.48 GB Validation, 4.85 GB Test.

Associated Challenges

There are four associated challenges with this dataset:

• Object Detection in Images
  
• Object Detection in Videos

  This challenge focuses on predict the bounding boxes with real-valued confidences of each object class in each video frames. Similar to the Object Detection in Image Challenge, we focus on 10 object categories of interest: pedestrian, person, car, van, bus, truck, motor, bicycle, awning-tricycle and tricycle. Some rarely occurring special vehicles (e.g., machineshop truck, forklift truck, and tanker) are ignored in evaluation.

  
• Single-Object Tracking

  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. The objects to be tracked are of various types including pedestrians, cars, buses, and animals.

  
• Multi-Object Tracking Evaluation

  This challenge focuses on 10 object categories of interest including pedestrian, person, car, van, bus, truck, motor, bicycle, awning-tricycle, and tricycle. Some rarely occurring special vehicles (e.g., machineshop truck, forklift truck, and tanker) are ignored in evaluation. The challenge will provide 96 challenging sequences, including 56 video sequences for training (24,201 frames in total), 7 sequences for validation (2,819 frames in total) and 33 sequences for testing (12,968 frames in total), which are available on the download page.

  

Licence

This dataset is made available for academic use only.