Overview

Latest updates
  • Apr 25, 2020: web site of the challenge opens, the task is revealed,
  • May 1, 2020: Training Data - releasing date has been postponed to May 10, 2020 due to the ongoing pandemic COVID'19!
  • Sept 02, 2020: The article explaining the India-Japan-Czech datasets and our analysis for the same, is now available(https://arxiv.org/abs/2008.13101).
  • Sept 10, 2020: Test2 has been released (https://github.com/sekilab/RoadDamageDetector). the submission form for test2 is opened. The final ranking is determined by the average scores of test1 and test2.
Description

Since roads have a direct and significant impact on people's lives, maintenance and management of roads need to be done exhaustively from time to time. However, the traditional methods for road condition assessment involves manual inspection, and are time-consuming, cost-intensive, highly subjective, and prone to errors. The problem is aggravated as the number of experts that can assess such road damages and make a decision regarding optimal resource allocation for repair works is limited.

Furthermore, a lack of financial resources makes many local governments unable to conduct sufficient inspections on time. Some municipalities automate road damage detection by using high-performance sensors. Nevertheless, the high cost of these sensors makes it infeasible to use them at the country level owing to the vast area of roads to be inspected.

Therefore, there arises a need for a system that makes it easy to assess the road conditions and identify the damage of the road surface at a low cost and in less time.

We have already developed a preliminary version of this system and also hosted a challenge in 2018 (IEEE BigData Cup) to evaluate the contemporary methods working towards the same goal.

After the road damage detection challenge in 2018, several municipalities in Japan started utilizing our automatic road damage detection systems. Through the practical use and from the feedbacks of several government agencies, we realized that the algorithms need to be more robust and versatile, especially in a real-world scenario. That is, the algorithm is required to perform well in a variety of situations, including the presence of shadows or reflections in the images. Further, it has been observed that most of the existing models are limited to road conditions in a single country. Proposing a method that applies to more than one country leads to the possibility of designing a stand-alone system for road damage detection all over the world.

Compared to the dataset in 2018, this year, we have increased the dataset volume three times. The challenge is to detect and identify the damages contained in road images captured by a vehicle-mounted smartphone. Since our dataset includes hard negative and positive examples collected in real-world situations from three different countries, the algorithm needs more robustness to meet the challenge.

Successful implementation of this challenge would open new horizons of possibilities where just the smartphones and drive cameras would be enough for road inspections, not for a single country but all the countries across the world.

The Task

Participants need to propose an algorithm that can automatically recognize the road damages present in an image captured from any of the following three countries: India, Japan, and Czech. The recognition implies the detection of the damage location in the image and the identification of the damage type.

Using this algorithm, the participants are required to predict labels for the test data. The predicted labels need to be submitted as a csv file through the submission link provided on the website.

After the competition phase is completed, a link for the submission of the accompanying academic paper will be provided to the top 10 participants as ranked by the public/private leaderboard weighting. For more details, please refer to the Submissions section.

Important Dates

This competition will continue to be on the site.
Schedule of Data Release:
  • Release of Training Dataset - May 10th
  • Release Test1 – May 15th
  • Release Test2 – Sept 10th
Important dates:
  • Apr 25, 2020: Website of the challenge opens, the task is revealed
  • May 10, 2020: Training Data - releasing date has been postponed to May 10, 2020 due to the ongoing pandemic COVID'19!
    Till then, interested candidates can explore the previously uploaded datasets on our GitHub page to have a glimpse of the data similar to the one to be used for this competition.
  • Sept 25, 2020: Deadline for submitting the source code & the solutions, End of the competition
  • Oct 5, 2020: Announcement of winning teams, Sending invitations for submitting papers for the special track at the IEEE BigData 2020 conference
  • Oct 25, 2020: Deadline for submitting invited papers,
  • Nov 10, 2020: Notification of paper acceptance,
  • Nov 20, 2020: Camera-ready of accepted papers due,
  • Dec 10-13, 2020: The IEEE BigData 2020 conference (special track date TBA)