Overview

The structure of the dataset is illustrated below:

Several source videos have been split up into tracks. The tracks are annotated with the sign type they contain, but may include other signs. From each track, up to 30 frames are extracted and all signs in these frames are tagged with position, type, and some additional meta data.

Contents

• 47 US sign types
• 7855 annotations on 6610 frames.
• Sign sizes from 6x6 to 167x168 pixels.
• Images obtained from different cameras. Image sizes vary from 640x480 to 1024x522 pixels.
• Some images in color and some in grayscale.
• Full version of the dataset includes videos for all annotated signs.
• Each sign is annotated with sign type, position, size, occluded (yes/no), on side road (yes/no).
• All annotations are save in plain text .csv-files.
• Includes a set of Python tools to handle the annotations and easily extract relevant signs from the dataset.

Citation

Please use the following citation when referencing the dataset:

@article{mogelmose2012vision,
  title={Vision-based traffic sign detection and analysis for intelligent driver assistance
systems: Perspectives and survey},
  author={Mogelmose, Andreas and Trivedi, Mohan Manubhai and Moeslund, Thomas B},
  journal={IEEE Transactions on Intelligent Transportation Systems},
  volume={13},
  number={4},
  pages={1484--1497},
  year={2012},
  publisher={IEEE}
}

License

Custom