• There is no restriction on the type of algorithm/model you are using. Any algorithm, irrespective of whether it utilizes deep learning or not, is allowed.
  • Pre-trained models are allowed in the competition.
  • Participants are restricted to train their algorithms on the provided Road Data train sets. Collecting additional data for the target attribute labels is not allowed. If you do so, please mail to the organizers ( However, it is permitted to increase training images using data augmentation, GANs etc. artificially.
  • Images on this dataset are available under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0). Creative Commons License
  • We expect you to respect the spirit of the competition and do not cheat. Hand-labeling is forbidden.
  • There is no restriction on the size of the team.
  • Multiple submissions are allowed, and a team can submit up to five csv files every day for judging.
  • If the performance of the algorithms submitted by two different groups is found to be the same, the one who submitted earlier would be ranked higher.
  • The top 10 rankers would be invited to submit the paper.
  • Please note that the ranks would be purely based on the performance of your algorithms, and the technical reports would not affect the ranking decision.
  • Even if you are submitting a paper with similar content to other workshops, if you add a 30% extension, you can submit through our link. However, we expect you to mention this point at the time of submission exclusively.
  • Please submit source code and trained model in executable form before the deadline. IPython Notebook is desirable for the source code submission, but any programming languages are acceptable. The organizers will verify the reproducibility of the algorithm and determine the final winner.
  • The prizes (as specified in the Prize section) will be offered to the top 3 rankers after the successful presentation of their paper at the conference.
  • There would be one special award for the team presenting the content, which can contribute to the future of road damage detection, such as suggestions for improving our dataset.