SVEN CV - the computer vision software used for SVEN (Surveillance Video
Entertainment Network http://deprogramming.us/sven) - is now available to
the public. The features and interface were designed specifically for the
SVEN project, so the interface isn't what you might expect from
user-friendly, general purpose software. However, we hope it can be useful
for other public space projects, as it is specifically designed to track
people in uncontrolled settings, as opposed to a gallery or stage where
lighting, background, clothing, etc., can be controlled.
About SVEN CV...
SVEN CV is a software application for real time tracking of pedestrians,
using OpenCV. Its robust build has features for tracking several
individuals with all kinds of foreground and background occlusions. It
also provides real-time subjective features such as face detection and
expression; hair and clothing color; segmentation of the body to give
positions for head, shoulders, and torso; direction of movement and more.
It transfers the data including the coordinates of the person's outline
(matte) to a specified IP address or the local machine via a UDP port. It
supports input from both live camera and AVI files.
There are already several computer vision applications available geared
toward "artists." So why did we bother to develop our own?
1) We wanted some specific features that weren't available in existing
software. These may or may not be of use to you. But more importantly...
2) We wanted it to work in public space. The SVEN project deals with
surveillance, and surveillance happens in public places: corridors,
storefronts, outside on the street, etc. The computer vision applications
for artists we found tended to assume a controlled setting like a gallery
or stage, where changes in lighting and background wouldn't be a problem.
While the best results with SVEN CV are still obtained with minimal
background changes and reasonably flat, constant lighting, SVEN CV is
designed to adapt to changes in background and applies heuristics to
eliminate as much as possible errors due to shadows and lighting changes.
Complete licensing information is included in the Readme. But in general,
SVEN CV and its source code can be used freely for educational, research
and non-profit purposes; however permission must be obtained for
SVEN CV software is written by Vincent Rabaud, developed in coordination
with Amy Alexander
The SVEN project was initiated and directed by Amy Alexander; Lead Video
Developer Wojciech Kosma; Lead
Computer Vision Developer Vincent Rabaud; Production Assistan Marilia
Maschion. Technical direction,
utility programming and occasional video programming by Amy Alexander.
Additional video programming by Jesse Gilbert; additional computer vision
programming by Nikhil Rasiwasia; additional utility programming by Cristyn
Magnus; additional production assistance by Annina Rüst.
Note - Mail sent to the email address in the header may or may not actually
reach me! To reach me at a current, fully functional address, please use the form on http://amy-alexander.com/contact/index.php. Gracias, danke, merci, and thank you!