I'm looking for annotated dataset of traffic signs. I was able to find Belgium, German and many more traffic signs datasets. The only problem is these datasets contain only cropped images, like this:

enter image description here

While i need (for YOLO -- You Only Look Once network architecture) not-cropped images.

enter image description here

I've been looking for hours but didn't find dataset like this. Does anybody know about this kind of annotated dataset ?


I prefer European datasets.


Direct Answer

The Belgium TS Dataset may be helpful, as well as The German Traffic Sign Detection Benchmark.

Additional Notes Based on Question Author's Idea

The idea in the question author's addendum of placing signs onto street sides and corners is a good one, but to do it repeatably and in a way that doesn't bias the training is its own research project. However, it is a good research direction. What would be of benefit to AV researchers worldwide is a multi-network topology and equilibrium strategy with the objective to create the following data generation features.

  • Street sign symbol inputs in image form, with or without cropping, as movie frame sequences or single still shots, or from SVG files.
  • Annotation generation using partially human-labelled data.
  • 3D analysis of sign angle and perspective setting so that the images appear exactly as they would from a vehicle's imaging system.
  • Matching of lighting between the superimposed sign and the background scene.
  • Automatic blue-screening for the sign image.

This is obviously not a basic data hygiene problem. It is its own AI project, but the return on this research project in terms of furthering the AV technology is immense and may have significant data set statistical advantages over collecting data from the vendors that supply images to Google maps and other Big Data aggregators.

  • 1
    $\begingroup$ I thought the German/Belgium dataset contains only cropped traffic signs, but recently i found out there are also datasets for detection purposes. Thank you for your answer. I think these datasets are going to do the work. $\endgroup$ – kocica Oct 21 '18 at 10:49
  • 1
    $\begingroup$ Really appreciate additional notes @Douglas Daseeco. Definitely you have got the point. The 3D analysis and matching of lighting came also to my mind. As you said, it's best for later down the road. Now I will try to train models on datasets mentioned in your answer and hope the results will be good. Then i would like to work on this artifical dataset generator and compare both models -- one trained on real dataset, another one on artifical. $\endgroup$ – kocica Oct 21 '18 at 15:14

Look at Google's Open Image Dataset @ https://storage.googleapis.com/openimages/web/index.html

They provide image-level labels, object bounding boxes, object segmentation masks, and visual relationships.

Here is the link for the traffic signs dataset.


Check this one by UCSD. It contains both video as well as images related to traffic signs. The annotations are present in csv

  • $\begingroup$ It seems like great annotated dataset. Actually i forgot to mention i prefer european traffic signs. Still its better than nothing, thank you. $\endgroup$ – kocica Oct 5 '18 at 8:21
  • $\begingroup$ My bad, I assumed you were looking for the US signs. Let me check if I can find European ones. $\endgroup$ – fireboy Oct 5 '18 at 20:31
  • $\begingroup$ Actually I've made simple "dataset generator", which places cropped signs into images of road, on the positions where signs usually are placed. It also create annotation file at the same time. You may check it here: github.com/kocica/training-dataset-generator, in the results folder. The results of neural network learned on this pseudo-dataset are surprisingly pretty well. $\endgroup$ – kocica Oct 7 '18 at 17:33

I searched the web but there are no such dataset published but Check this out


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.