DThree19

Is your deep detector safe? verification and certification of neural networks

CVPR Workshop, 17 June 2019, Long Beach, CA, US

Challenge

Visteon Lane Detection Challenge (ViLD) is about deployable and dependable lane detectors in the context of autonomous driving. This is a computer vision based challenge, organized and funded by Visteon Corporation . Unlike many other detection challenges, we focus on a very specific task, viz., detection of lanes painted on the road through front-mounted camera under variety of conditions (brightness, colour, curvature). Additionally, we are interested in real-world challenges of such a detection task. Therefore, instead of focusing on metric for quality of such detections, we ask ourselves how deployable these detectors are (i.e., can they run on impoverished hardware) and how dependable they are (i.e., would they perform well even when the environment changes slightly). For ease of evaluations, we assume that these solutions involve deep learning based approaches.

Please visit the website for the challenge: Visteon Lane Detection Challenge

Important dates

  • Phase-I ends: 31 March, 2019
  • Phase-II ends: 30 April, 2019
  • Winners announced: 17 June, 2019

Acknowledgment

We would like to acknowledge additional people behind the organisation of this challenge. ViLD has been actively supported by Anatoliy Atanasov, Ivaylo Stoyanov, Simeon Aleksandrov, Petar Verotnikov, Lyubomir Todorov and Manoj Sehgal.