Datasets for tracking

last modified : 19-07-2019

Here are listed all the datasets that can be used for tracking. Each dataset should come with a small description of its size, what's in it and who provided it.

The UA-DETRAC Benchmark Suite

This dataset is both for multi-object detection and multi-object tracking.

• Contains:

  • Vehicles: Cars, Bus, Van, Other
  • Weather: cloudy, sunny, rainy, night
  • Different level of occlusion

• Size:

  • more than 140 thousand frames
  • 8250 vehicles manually annotated
  • 1.21 million labeled bounding boxes of objects

• Other details:

  • location: 24 different locations at Beijing and Tianjin in China
  • 10 hours of videos captured with a Cannon EOS 550D camera

• Article:

  • Title: UA-DETRAC: A New Benchmark and Protocol for Multi-Object Detection and Tracking
  • Authors: Longyin Wen, Dawei Du, Zhaowei Cai, Zhen Lei, Ming{-}Ching Chang, Honggang Qi, Jongwoo Lim, Ming{-}Hsuan Yang and Siwei Lyu
  • Link: article

• Dataset: here

AI City Challenge (2019)

This dataset is linked to a challenge, it is unsure whether they will give the data if you do not participate in the challenge. Also the data may change from year to year as the challenge evolves

• Contains:

  • Vehicles, urban objects (may not be labeled)
  • Metadata of the cameras

• Size:

  • 25h of videos for the highways
  • 3 hours of synchronized multiple cameras for the city

• Other details:

  • location: Highways in Iowa, Images from city in the USA

• Dataset: here