State-of-the-Art Technology

Building on over 10 years of R&D, Pittsburgh Pattern Recognition is pioneering state-of-the-art technology in automated video analysis. Historically, it has proven difficult to achieve automated video analysis at the level of human performance due to variations in object appearance (e.g., pose, scene lighting, facial expression) as well as variations in video characteristics (e.g., low-resolution/poor video quality, motion blur). PittPatt's technology is overcoming these barriers through a robust machine-learning approach. This approach minimizes modeling assumptions by learning all aspects of object models over massive amounts of real-world data. Our universal classification infrastructure supports three fundamental components:

• Object detection: automatically finding and locating objects that belong to a given category, such as faces or vehicles.
  
• Object tracking: automatically tracking objects in video by associating unique identifiers to individual objects across time.
  
• Object recognition: automatically identifying specific instances of general object categories, as in face recognition.