Personal tools
  • Log in
You are here: Home Events Competition
Navigation
The ARTTS project is funded by the European Commission (contract no. IST-34107) within the Information Society Technologies (IST) priority of the 6th Framework Programme.
 
Document Actions

ARTTS Time-of-Flight Face Detection Competition

by Martin Haker last modified 2010-09-08 10:49

ARTTS Time-of-Flight Face Detection Competition

To spur progress in time-of-flight-based object detection, the ARTTS project has hosted a competition on face detection using time-of-flight (TOF) cameras. The prize for the competition was a MESA SR4000 camera.

We are proud to announce the winner of the ARTTS TOF face detection competition:


Thomas Zielke, Prof. Dr.-Ing. M.Sc.
Jens Lippel
André Stuhlsatz
Fachhochschule Düsseldorf
University of Applied Sciences


3dface_small.png sr4000.png
MESA SR4000

Participants were provided with a training dataset of face and nonface TOF images. The entries, which had to be submitted as a Linux or Windows executable, were tested on a dataset of new, unknown images. Each test image was a 24×24 pixel patch, and the detector had to output a real-valued decision value for each image; an ROC curve was computed from these decision values (for details, see the technical information, provided on request).

Because low false-positive rates are critical for face detectors, the detectors were only evaluated on the range of false-positive rates between 0 and 5·10-6. The entry that achieved the greatest area under the curve (AUC) on this range of false-positive rates won the prize; additionally, entries had to achieve an average detection rate over this range of at least 0.95 to qualify for the prize.

To obtain the training dataset and further technical information, write an email to info@artts.eu.

The sponsors of the competition are:


mesa_imaging.png

Mesa Imaging


artts_logo_h50.png

ARTTS / University of Lübeck, Germany


gnns.gif

German Chapter of the European Neural Network Society


smi_logo_small.png

SensoMotoric Instruments



prc_logo_small.png

The Pattern Recognition Company


The ARTTS project is funded by the European Commission (contract no. IST-34107) within the Information Society Technologies (IST) priority of the 6th Framework Programme. However, this does not constitute an official endorsement of this competition by the European Commission.


Competition rules:

  • The competition is organized by the Institute for Neuro- and Bioinformatics of the University of Lübeck, Germany.
  • Only one entry may be submitted per participant, though we will allow participants to replace their existing entry with a newer entry at any time before the entry deadline.
  • In the event of a draw, the prize will be awarded to the entry with the lowest running time.
  • Entries that require longer than one hour to classify 1 million test images on the test hardware (a 2.66 GHz Intel Core i7 with 4 GB of RAM) will be disqualified.
  • Entries must be submitted to the competition organizers by the entry deadline of midnight Coordinated Universal Time on February 28, 2010. In the event that technical problems for which the organizers are responsible prevent an entry from being received in time, the organizers reserve the right to accept entries after this date at their discretion; however, the organizers are not obliged to exercise this right.
  • The competition sponsors, the ARTTS consortium partners, their employees and their employees' relatives may not participate in this competition.
  • In the event of unforeseen circumstances that are not covered by the competition rules, the organizers will resolve the issue according to their best judgment. The organizers' decision is final.
  • No part of the competition invitation and rules shall be construed as a contract or part of a contract. No part of the competition invitation and rules shall create any legally binding obligation on the competition sponsors, the ARTTS consortium partners, the European Commission, or any other party.
 

Powered by Plone CMS, the Open Source Content Management System

This site conforms to the following standards: