German Traffic Sign Detection Competition

The German Traffic Sign Detection Benchmark is a multi-class detection problem in natural images. Competition results will be presented at the IJCNN 2013.

The Traffic Sign exampleGerman Traffic Sign Detection Competition has started!

Results will be presented at IJCNN'13 in Dallas Texas, and FREE REGISTRATION prizes are to be won!

Competition Design

The "German Traffic Sign Detection Benchmark" is a multi-class detection problem in natural images. We do cordially invite you to participate. Our benchmark has the following properties:

  • single-image detection problem
  • images with zero to four relevant objects
  • 900 images (divided in 600 training and 300 evaluation images)
  • division in three categories
  • industry-relevant application: traffic sign detection

The first competition information was released on December 1, 2012, when the preliminary training data was published. Interested teams are invited to download the training dataset and develop their algorithms based thereon. In February there will be a submission phase where teams are invited to upload their results on a new unannotated test dataset. It will then be possible to immediately compare all competing algorithms via our website.

The official URL to the benchmark is: