Watching People


VicarVision is involved in developing a broad range of technologies that enable computers to autonomously monitor indoor and outdoor environments. These technologies are utilized in a wide range of applications like consumer behavior analysis, surveillance and elderly care. For such challenging purposes, an algorithmic pipeline with real-time detection and tracking of people is required with maximum accuracy and minimum computational complexity.

People Detection

The wide application capabilities of real time monitoring systems create a new era for surveillance. Due to the vast amount of available content, scene understanding turns out to be a matter of not only accuracy but also complexity. We, at VicarVision, focus on creating local, embedded or cloud based monitoring systems with optimal accuracy and complexity. The proposed systems are applied to varying scenarios including but not limited to the following use cases:

  • Retail: People counting and heat mapping

    Market research for consumer analysis are used to resolve the motivations and implications underlying the shopping and consuming attitudes.
  • Behavior analysis: People tracking

    Tracking people traffic is crucial for generating the behavior patterns of the people with respect to locational and time based services.
  • Surveillance: People detection & tracking

    People surveillance is not only a technical but also a psychological challenge. With such a powerful tool, suspicious and unexpected behavior is detected before it is too late.
  • Elderly care: detecting falls, detection of wandering elderly

    Elderly people could make use of autonomous care services where health related hazardous occasions could be informed to the attention of medical authorities.
Video demonstrating our People Detection technology.

Body Pose Estimation

Besides the current available technologies on ‘watching people’, VicarVision is doing extensive research in topics like body pose estimation and human action recognition. Bodily movements could be beneficial for action classification; hence, a reliable pose estimation capability is vital in human monitoring systems.