Data has always been a decisive factor when developing algorithms and models in the field of AI. Realizing that, the VIA team considers open datasets a central part of the project. VIA datasets 1.0 is packaged and released to the community, including 2 small datasets:

• VIA Traffic sign dataset - Sign detection data for model autonomous vehicles

  • Image data taken from self-driving cars of the Digital Race 2020 - FPT.
  • 11095 photos contributed by 2 groups Goodgame and ICT.
  • 6 sign data classes: stop, left, right, straight, no_left, no_right.

• VIA Lane segmentation dataset - Line segmentation data

  • Image data taken from self-driving cars of the Digital Race 2020 - FPT.
  • 7688 photos collected and labeled by the Goodgame team.
  • 3 data layers: background, road, lane_line (road markings).

This dataset will be the base dataset to build deep learning models for VIA project in the near future. We have also made it public for the community to download and use for free.

Download this dataset here: https://github.com/makerhanoi/via-datasets/releases/tag/v1.0.