The DARPA Neovision2 project aims to create neuroscience-inspired visual systems that can detect, recognize, and track many different classes of objects in live video imagery. To exercise the systems, Neovision2 teams collected and annotated a two unique sets of 100 high-definition video clips each:
Human annotations (ground truth) are provided every 4 video frames for 10 object categories: Boat, Car, Container, Cyclist, Helicopter, Person, Plane, Tractor-Trailer, and Truck. Developers of machine vision algorithms can use these annotations to compare the objects detected by their algorithms to those detected by human operators.
The following documents provide more information about the datasets:
You may also be interested in the Neovision2 evaluation software, which can score your algorithm output against the ground truth data: NeoVision2Eval-2.2.zip (1.3 MB). This software has been created by SET Corporation, an SAIC company.
The datasets and the above documents have been approved for public release (DISTAR case 21306).
This research project was made possible by funding from the Defense Advanced Research Projects Agency. The authors of this document affirm that the views, opinions, and data provided herein are solely their own and do not represent the views of the United States government or any agency thereof.
Copyright © 2013 by the University of Southern California, iLab and Prof. Laurent Itti