Corner cases represent unexpected and potentially dangerous situations, and the ability to describe them is not easy, but brings benefits. For example, corner case description can be used to generate synthetic corner case data for model training or to specify test cases for model validation. The goals and challenges associated with corner case description can be found in the paper "Description of Corner Cases in Automated Driving: Goals and Challenges" by Daniel Bogdoll, Jasmin Breitenstein, Florian Heidecker, Maarten Bieshaar, Bernhard Sick, Tim Fingscheidt, J. Marius Zöllner and will be presented at the Embedded and Real-World Computer Vision in Autonomous Driving (ERCVAD) Workshop at ICCV 2021.

Paper preprint: https://arxiv.org/abs/2109.09607

This work was carried out as part of the AI Data Tooling project (part of the AI family).