Hiwi, Bachelor or Masterar Thesis: Crowd Monitoring in Aerial Imagery

The Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB is one of the biggest institutes for applied research in the field of image acquisition and exploitation in Europe. The department Video Exploitation Systems (VID) devotes itself to automatical signal processing of moving imagery sensors in complex and non-cooperative scenarios. Such sensors are applied within the fields of reconnaissance or surveillance as integrated components in aerial, spaceborn or land-based platforms. VID develops and deploys computer vision algorithms for autonomous and interactive systems.

Figure 1:  Wen et al. „Drone-based Joint Density Map Estimation, Localization and Tracking with Space-Time Multi-Scale Attention Network“



Crowd monitoring is an important task for planning and conducting events such as public festivals, as e.g. the Hamburg Port Anniversairy, the Cannstatter Volksfest or Das Fest in Karlsruhe. To assure the safety of visitors, organizers typically perform simulations before the actual event takes place. However, to evaluate the performance of such a simulation the real behavior of people has to be measured.


Your task

For this work you will do an extensive literature research on video-based crowd counting algorithms. Based on your results you will choose to create an own approach for tackling the underlaying problem. The focus will lay on incorporating temporal information to achieve a robust and accurate crowd count estimate from aerial imagery.



  • You study: Computer Science, Mathematics, Electrical Engineering, Applied Physics or comparable
  • Good programming skills (ideally Python and C++)
  • Experience with deep learning frameworks such as PyTorch
  • Ability to work independently
  • Willingness and joy to explore new fields on your own and to bring your ideas to life


[1] Bahmanyar et al.: “MRCNet: Crowd Counting and Density Map Estimation in Aerial and Ground Imagery“ https://arxiv.org/abs/1909.12743
[2] Wen et al.: “Drone-based Joint Density Map Estimation, Localization and Tracking with Space-Time Multi-Scale Attention Networkhttp://arxiv.org/abs/1912.01811
[3] Overview on Crowd Counting Algorithms: https://github.com/gjy3035/Awesome-Crowd-Counting


Please send your application to Thomas Golda.