Tracking People in Aerial Image Sequences

Beschreibung:

The datasets above have been created for the research project VABENE - People Traing in Aerial Image Sequences. Each dataset comprises of an aerial image sequence and a xml file with manually labeled trajectories of all visible persons.

 

Many thanks to the German Aerospace Center for funding the project and providing the aerial image sequences.

DLR-Logo

Changelog

2012/10/26
The results of the IPF detection algorithm for single persons in aerial images are available for download. They can be used e.g. to evaluate and compare different tracking algorithms.

2012/10/01
XML files have been updated. Every objects is now described by an image representation in addition to its bounding box. Furthermore we provide C++ code, so you can easily import and use the data in your project.

Training Datasets

 

Name Characteristics Data
Munich01
  • 24 frames
  • ca. 2.5Hz
  • pixel size 12cm
  • 55 individuals / frame
AA_Crossing_01
  • 18 frames
  • 2Hz
  • pixel size 15cm
  • 150 individuals / frame
AA_Easy_01
  • 14 frames
  • 2Hz
  • pixel size 15cm
  • 8 individuals / frame
AA_Easy_02
  • 12 frames
  • 2Hz
  • pixel size 15cm
  • 15 individuals / frame
AA_Easy_Entrance
  • 19 frames
  • 2Hz
  • pixel size 15cm
  • 60 individuals / frame
AA_Walking_01
  • 13 frames
  • 2Hz
  • pixel size 15cm
  • 35 individuals / frame
RaR_Snack_Zone_01
  • 4 frames
  • 2Hz
  • pixel size 15cm
  • 230 individuals / frame

 

Evaluation Datasets

 

Name Characteristics Data
Munich02
(AIS3)
  • 31 frames
  • ca. 2.5Hz
  • pixel size 12cm
  • 200 individuals / frame
AA_Entrance_01
AA_Entrance_01
  • 16 frames
  • 2Hz
  • pixel size 15cm
  • 870 individuals / frame
AA_Entrance_01
AA_Crossing_02
(AIS1)
  • 13 frames
  • 2Hz
  • pixel size 15cm
  • 90 individuals / frame
AA_Walking_02
(AIS2)
  • 17 frames
  • 2Hz
  • pixel size 15cm
  • 160 individuals / frame
RaR_Snack_Zone_02
  • 4 frames
  • 2Hz
  • pixel size 15cm
  • 220 individuals / frame
RaR_Snack_Zone_04
  • 4 frames
  • 2Hz
  • pixel size 15cm
  • 310 individuals / frame