Image database for vehicle recognition


The dataset is composed of videos containing different types of vehicles travelling on diverse kind of roads, normally highways. Unless the contrary is explicitly said, all videos are in RGB. Three sources have been used to compose this dataset, which are:

  • DGT Database: some traffic cameras publicly available from the DGT (Dirección General de Tráfico) have been used to record only video at night, as shown below in the "Samples" section.
  • Gelderland Database: a traffic camera publicly available from a road in Gelderland has been used to record video in both day and night. When it is night, this traffic camera provides monochrome video (gray scale), as shown in the "Samples" section.
  • UA-DETRAC Database: this database comes from the challenge proposed in the link provided below. Some of the included sequences have been used for tests, which in concrete are those in the link Image database for vehicle recognition, also provided below.

All sequences from the dataset, except from UA-DETRAC ones, were recorded with traffic cameras which introduce much compression in images, while UA-DETRAC sequences were recorded with a high-quality CANON camera. DGT and Gelderland databases have been gathered for the GTI specifically for the DGT Project, while UA-DETRAC Database does not come from the GTI, but it already exists publicly.

Types of vehicles

- Nighttime: Vehicle

- Daytime: Bus, Car, Lorry, Van, Bicycle, Motorbike, Car with trailer, Lorry with trailer, Van with trailer

Samples of each scenario of the considered roads

  • DGT Database:

  • Gelderland database:

  • DETRAC database: