Video Analytics Technology Evaluation


IEEE Change Detection Evaluation

Proving our robust video analytics capabilities, intuVision participated in the IEEE Workshop on Change Detection. We submitted our algorithms against 21 other teams from universities and research centers around the world, and received the top score in both the dynamic background and thermal categories, and second in the shadow category. These rankings were determined using standard measures such as F-measure, precision and recall.

This was an independent evaluation designed to judge the current state of video analytics algorithms from research centers around the world. Participants submitted algorithms designed to detect various objects in many different environments, rewarding those that return good detections in diverse settings.

Out of all of the participants, intuVision provides the only algorithms for purchase. These top ranked algorithms are the same that power our video analytics solutions.

Example shadow detection

Example shadow detection. Left to right: Marked reflections; Ground truth foreground; Calculated background model; Foreground result.

Read report summary >>

NIST TRECVid Event Detection Evaluation

Proving the strength and capabilities of our video analytics technology and expertise, intuVision participated and received top rank in TRECVid. TRECVid is an evaluation system developed to give video analytics organizations a baseline upon which to compare products. It consists of a series of tasks which companies then modify or expand their current products in order to complete.

 

The evaluation in which intuVision participated involved analyzing several hours of surveillance video data from a major airport for three event detection scenarios, described below. Building upon our expertise of the subject and the video analytics algorithms that comprise Panoptes, intuVision video analytics were able to intelligently detect all three events.

Detection Scenarios:

  • Opposing Flow: Someone enters through the Airport Terminal No Re-Entry doors.


  • There are four images from surveillance video.  The first, and largest, is of an airport, with everything blurred out except for a rectangle around the top of the door, showing the area to be analyzed. The next three (smaller) images show a man walking through the door, and an alarm being triggered when he is through the door.
  • Elevator No-Entry Event: A person waits for the elevator, but does not enter when the elevator door opens.


  • Shows four images from an analyzed surveillance video, in the first three there is a man waiting outside of the elevator outlined in a green box. In the fourth the elevator doors are open, but the man has not gone and is in a red box.  This fourth one is the alarm frame.
  • Take Picture: Someone takes a picture in the monitored area. The alarm frames are the frame in which the person first starts taking the picture and the frame in which the flash occurs.


  • Shows two images of a crowded airport lobby.  In the first a person is beginhing to lift their arms to take a picture. In the second, the person is taking a picture, and you can see the flash.  In both the person is outlined in a red box and under the images it says alarm frame.

Read evaluation report >>

Products Technology Partners About Us News Contact
Panoptes
Solutions
About our Partners About intuVision News Contact Us
VideoRecall R&D
Become a Partner Career Opportunities Press Releases Become a Facebook Fan
intuVision SDK
Publications   Contact Us eNewsletters Follow Us on Twitter
Customer Case
Evaluations Directions Test Our Products