ABOUT

Visteon is a leading global supplier of automotive electronics and the only one focused exclusively on cockpit electronics. The company design, develop and produce complex technology products such as Driver Information Clusters, Displays, Connectivity Management Units and Telematic Modules. Our solutions guarantee maximum comfort for drivers and passengers while staying connected. The latest development areas we focus on are ADAS (Advanced Driver Assistance Systems) – the first mass adopted solutions that will pave the road for the autonomous driving.

Going along with this trend and following the strategy of the company’s leading R&D center, Visteon Bulgaria is hosting the first National Automotive Hackathon. Our goal is to find more people like us – passionate about developing intelligent and safe systems in the car. Let’s turn ideas into real solutions and why not you to become part of our teams for a longer journey in the automotive world!

CHALLENGES

1

Visual Odometry

The Visual Odometry Challenge (Challenge #1)

Based on a sequence of images your solution should be able to predict the relative position of a vehicle.

Reference: Haarnoja et al., “Backprop KF: Learning Discriminative Deterministic State Estimators”, 2017

A reference paper is available here: https://arxiv.org/abs/1605.07148

Evaluation will be done using the translation metric defined here:
http://www.cvlibs.net/datasets/kitti/eval_odometry.php

The requirement for all performance rewards is maximum error of 2%

Additional sample data and evaluation will be available starting Nov. 2019

2

Visual Depth Perception

The Visual Depth Perception Challenge (Challenge #2)

Based on a sequence of images your solution should be able to predict the depth map associated with each image.

Reference: Casser et al., “Depth Prediction Without the Sensors: Leveraging Structure for Unsupervised Learning from Monocular Videos”, 2018

A reference paper is available here: https://arxiv.org/abs/1811.06152

Evaluation will be done using the SILog metric defined here:
http://www.cvlibs.net/datasets/kitti/eval_depth.php?benchmark=depth_prediction

The requirement for all performance rewards maximum error: 18.0

Additional sample data and evaluation will be available starting Nov. 2019

Your ideas may save lives!

RULES

 Sign up by sending E-mail to compete@visteon.com containing the following information:

  • Participant/Team nickname
  • Valid GitHub account name
  • Statement acknowledging that you agree to the terms and conditions

We have set up two challenges and support data for finding out who can design solutions with high performance and accuracy. Starting in October 2019 for 4 months we will look for and support the development of algorithms with open source code that are mature and secure enough to be used in the AI systems of future vehicles. After sending your registration request you will be granted access to a repository where you will upload your programs. Every submission in GitHub will be run against benchmark solution, several mentors will answer implementation questions and the best ones will bring a prize to their creators.  

Submissions must be able to compile and run on an Ubuntu 18.04.3 LTS (Bionic Beaver) amd64 machine. Participants may submit an install script (to be run without root privileges) or list of Ubuntu packages that can be installed to facilitate testing.

If CUDA/OpenCL is applicable or used it must be compatible with a GeForce GTX 1050 GPU. Further restrictions on the CUDA/OpenCL version may be placed during the competition.

Starting in November 2019, submissions will be evaluated and results will be available for each week for any submission submitted by Monday 00:00 GMT.

Submissions will be tested on internal data. Samples of internal data will be supplied during the competition.

  • The participant must ensure that the code being supplied is commercially usable and properly licensed according to the competition rules. The following license is an example that will be acceptable if you include it in a LICENSE.txt file accompanying your submission:

Copyright (C) <2019> <Visteon Corporation>Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby granted.THE SOFTWARE IS PROVIDED “AS IS” AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.

 

  • Including other and/or alternative licenses may disqualify your submission.

PRIZES

Most

accurate solution

500 Euro

+200 EUR bonus if the solution is written entirely in C++14, CUDA and/or OpenCL.

Solution with

best performance

300 Euro

Must meet a minimal accuracy level defined for each challenge separately.

Best performing solution

written entirely in C++14

500 Euro

Must meet a minimal accuracy level defined for each challenge separately.

CUDA/OpenCL NOT accepted for this reward.

 

COMMUNICATION

Participants may submit a text file containing questions they have accumulated during the week.

Each Monday questions from the previous week will be answered on the AI Challenge webpage under the FAQ section.

Starting November 2019, for each challenge, a Microsoft teams chat will be available each Monday from 10:00-12:00 GMT for an interactive question and answer period.

A leaderboard will be available for each reward on the AI Challenge webpage.

MORE

The VISTEON ADAS AI Challenge is open to all researchers, engineers and enthusiasts.

It is being organized by the VISTEON ADAS team in Sofia, Bulgaria.

The intent is to give practitioners around the world a venue in which to develop their software skills and understanding of ADAS and related machine learning topics.

Multiple rewards totaling up to 3000EUR will be awarded to the best solutions based on different criteria.

Throughout the challenge participants will have access to Q&A by professionals in the ADAS Perception team.