Edge AnnotationZ Challenge

AI Sweden, Zenseact, RISE, Lund University, Chalmers Industriteknik, and Univrses, in collaboration with CGit and MobilityXLab, invited the AI community to work on one of the key challenges of autonomous driving and contribute to the development of future road safety.

Application deadline: Closed
Challenge: Ended

Urban traffic scene with a pedestrian crossing, viewed from the perspective of a driver

Image by Zenseact

Zenseact is looking for ideas and solutions to tackle the obstacles of automatic annotation. Seize the opportunity to partake in this challenge to accelerate the development of autonomous driving and impact the future of mobility. The Edge AnnotationZ Challenge is part of the Road Data Lab project. Zenseact is interested in potentially continuing collaboration with the top solutions of this challenge.

The Challenge

The automotive industry stands on the verge of a great revolution in autonomous driving, the Edge AnnotationZ Challenge will be an exciting experience to develop new solutions for high-quality automatic annotation generation, making edge learning a reality on a larger scale! We hope to develop solutions that can enable vehicles to improve their perception systems without any human supervision.

Download the complete problem formulation

For participants & prize

1. Solved a key challenge together with researchers from Zenseact.
2. Received access to a world-class road data set and got to use the powerful infrastructure of the Data Factory (NVIDIA DGX A100).
3. A price sum of 50,000 SEK (approximately 5,000 EUR) and a chance of further collaboration with Zenseact.

What's the Road Data Lab (RoDL) project?

This Challenge is part of the Road Data Lab (RoDL), an initiative from RISE to establish a platform for technical infrastructure, legal frameworks and multiple sources of road data to further advance collaboration, learning and innovation in regards to road data.

Legal Notice
1. The dataset provided by Zenseact is the property of Zenseact AB (© 2021 Zenseact AB)
2. The dataset is licensed under CC BY-SA 4.0.
3. Any public use, distribution, or display of this dataset must contain this entire notice:
For this dataset, Zenseact AB has taken all reasonable measures to remove all personally identifiable information, including faces and license plates. To the extent that you like to request the removal of specific images from the dataset, please contact privacy@zenseact.com.

Evaluation of results

The solution was evaluated based on how good edge annotations you provided at the target frame. The metric for determining the quality of your annotations was mean average precision. It was noted that the mean average precision was a (weighted) average over the 2D and 3D properties.

As mentioned previously, the provided code had to be runnable on the AI Sweden infrastructure environment to be considered for evaluation. Failing to comply with this led to an invalid result. Moreover, the solution had to comply with the spirit of the challenge for eligibility.