Our novel visual analytics framework is based on a service-oriented processing paradigm permitting optimal data, task and computational load distribution within a network of connected devices. Visual analysis, data handling and visualization are encapsulated within respective modules which are self-describing and work as scalable services on their own. Requirements of interconnectivity across various devices of heterogeneous character and scalability can be easily met: data are described in a light-weight data interchange format (JSON), while data handling, buffering and the underlying service framework are generated in an automated manner.

Application Example: Interactive Video Exploration

Visual information in digital form is everywhere. Various authorities such as infrastructure operators and law enforcement agencies are confronted with search needs based on a visual description. Technological advances by steadily improving cameras, progress in Computer Vision, increasing computational power and the wide availability of network-based connectivity between devices are opening new horizons to tackle the task of finding the needle in a haystack of digital data.

Several algorithmic components (e.g. detection, tracking, license plate recognition) have been combined and integrated into an interactive tool which provides search capabilities across various attributes and in joint feature spaces in large visual surveillance archives. This tool provides a convenient means to find specific video objects (such as a person or vehicle) in vast sets of video data. The system was awarded with the Best Demo Award from IBM at the 2011 IEEE Int. Conf. on Multimedia & Expo (ICME) in Barcelona.

Technology platform:

  • Video Archive Search Manager - a prototype tool to look for specific events (movements, Persons, Similarities, etc.)

Reference projects:

  • SECRET (Search for Critical Events in Video Archives)
  • SECRET II (Search for Critical Events in Video Archives) - Continuation of previous project.
  • CriFi (Critical Content Filtering)

Industrial Reference projects:

  • Siemens AG: People re-identification and object detection algorithms

Demo Video - Video Archive Search by Avatar Modelling

Demo Video - Video Archive Search by Video Snapshot