Skip to content
Branch: master
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


For the Cifar10 and MNIST networks, pylearn2 to is used to provide the training data. For the GTSRB example, we need to download the GTSRB dataset from Furthermore, for the GTSRB example, we also provide a "junk" class, which is used to avoid false positives during inference, for non-traffic sign related images.

German Traffic Sign Detection Benchmark Junk Class

We also provide an extra set of images to be used as an extra output in the network. As stated above, this is to avoid false positives during inference.

This dataset, comprises of images from GTSDB which has been broken into small patches. A subset of these patch were then chosen which had:

  • no traffic signs in them; and
  • as much variation as possible (e.g., trees, road, sky and cars).

The variation in the chosen patches was done to ensure the inference models were as resilient as possible to classifying non-traffic sign related images.


Thanks to the Institut Für Neuroinformatik Benchmark Team for allowing us to provide you with this modified version of the GTSDB to provide background images.

If you plan to use this work, please cite the GTSDB paper, and explain the modifications stated above. The GTSDB citation is provided below:

   author = {Sebastian Houben and Johannes Stallkamp and Jan Salmen and Marc Schlipsing and Christian Igel},
   booktitle = {International Joint Conference on Neural Networks},
   title = {Detection of Traffic Signs in Real-World Images: The {G}erman {T}raffic {S}ign {D}etection {B}enchmark},
   number = {1288},
   year = {2013},
You can’t perform that action at this time.