Researchers from the Augmented Vision group at DFKI have developed a novel approach for automatic video-based traffic analysis.
First, entry and exit zones of complex crossroads or roundabouts are defined interactively.Then, each car is tracked individually by computing the trajectory based solely on the video content.
This provides unique data for traffic analysis and in particularfor supporting simulations for improved traffic solutions.

Feel free to check out the demonstration videos on our YouTube channel: Traffic Analysis

In this paper we present a one-button photo-realistic scanning system which provides a realistic 3D (usable!) representation of a given object. The paper will be presented at the Digital Heritage Conference from the 28th October – 1st November 2013 in Marseille, France. Feel free to check out the demonstration videos on our YouTube channel:

• 3D Reconstructions Representation
• OrCaM System demonstration

We are very proud to announce that we have won the Google Research Award for our research in 3D scanning, object reconstruction and appearance modelling. With this award Google supports cutting-edge research in Computer Science and Engineering with the aim of identifying and supporting world-class, full-time research of mutual interests. The award was granted to the Augmented Vision group for their outstanding work in new methods to fast and systematic 3D modelling and rendering of object appearance.For more information on the OrCaM-system, please visit our project page under http://av.dfki.de/orcam or check out our videos on YouTube:

• 3D Reconstructions Representation
• OrCaM System demonstration

In this paper we present a novel approach for a user supported tracking framework which combines automatic tracking and minimal user input for error free tracking results suitable for interactive video production. We present our results on the CVMP, the 10th European Conference on Visual Media Production on the 6-7th November 2013 in London. You can download the paper here or browse our list of publications.

Twelve different partners from industry and research meet on the 17 & 18th of October at the DFKI in Kaiserslautern for the kickoff meeting of the EASY-IMP project.

Scanning people with a Kinect is still a challenging problem. In this poster we demonstrate an approach for a fully automatic, full body scanning, combining state of the art super-resolution and registration algorithms with a novel approach for shape completion to reconstruct the human body.

For a copy of the poster, please contact Yan Cui.

Two papers have been accepted for the upcoming International Symposium on Wearable Computers (ISWC):

Attila Reiss, Gustaf Hendeby, Didier Stricker

Confidence-based Multiclass AdaBoost for Physical Activity Monitoring

IEEE 17th International Symposium on Wearable Computers (ISWC), September 9-12, 2013, Zurich, Switzerland.

Attila Reiss, Didier Stricker

Personalized Mobile Physical Activity Recognition

IEEE 17th International Symposium on Wearable Computers (ISWC), September 9-12, 2013, Zurich, Switzerland.

For the conference program and the papers’ abstract, please visit: http://www.iswc.net/iswc13/program.html.

The first two projects of the software-cluster EMERGENT and SWINNG have been successfully evaluated by DLR and BMBF. In addition to the generally positive review the BMBF reviewers chose a small selection of “gems”, among them AV’s AR-Handbook for their novel approach to Augmented Reality maintenance assistance.

In the next two years the German Research Center for Artificial Intelligence (DFKI) and  the Institute for Applied Scientific Training (IAT) will work together on the use of novel technologies in professional sports. German press release

Motion sensors are becoming incredibly cheap and are already fully integrated. Chaining them allows fully new applications in the area of body and tools tracking.

For a full copy of the work, please contact Markus Miezal.