AVA Challenge: Anticipating Vehicle Accidents Challenge.

Announcement of Challenge Winners

We're excited to reveal the top performers in the Anticipating Vehicle Accidents Challenge! The following participants will be awarded certificates, and the first-place winner will receive an additional prize. Congratulations!!!

1st Place: Matthew1012, with a score of 0.67142

Solution: Predicting Car Accidents with YOLOv7 Object Detection and Object Relationships

2nd Place: ALLAccept, with a score of 0.61428

Solution: Improved Dynamic Spatial-Temporal Attention Network for Early Anticipation of Traffic Accidents

News

The paper submission deadline has been moved up from **May 1** to **April 28** due to ICME committee decision. We will send out the acceptance notice of your papers on April 29. Participants are kindly reminded to take note of this change.

The valid submission only if the participants have submitted their report, code, and successful submission on kaggle could be considered as the winner. For the winners of the competition (top three), we will invite the top three to five to share their winning solutions at ICME and award certificates to the top three. In addition, our sponsors will provide an NVIDIA 3090 GPU as a prize for the winners.

Private Leaderboard Score Released

We are pleased to announce the top three performers on our private leaderboard for the recent competition. These participants have achieved remarkable scores.

1st Place: Matthew1012, 0.67142

2nd Place: Huang Yu Hung, 0.61904

3rd Place: ALLAccept, 0.61428

4th Place: asd3200asd, 0.59047

5th Place: Host(not included)

6th Place: abida gurman, 0.58571

We will send an email to the above participants to notify them to submit a link to their technical report (1 page) and source code. Please stay tuned for this important message.

All of the aforementioned participants are required to submit their challenge papers. Please note the deadline for paper submission and we encourage all participants to submit their work as well.

Once we have verified that your technical report and source code are correct, you will be qualified to submit a challenge paper. The award of the prize and certificate will depend on the quality of your submitted paper. Failure to submit a paper or submitting a paper less than 3 pages in length will result in the cancellation of the award.

(一旦我們驗證了你的技術報告和原始碼無誤，你就有資格提交競賽論文。獎品和證書的頒發將取決於你提交的論文品質。未提交論文或提交長度少於3頁的論文將取消你的 獎項。)

Organizers

• Prof. Chih-Chung Hsu, Assistant Professor, Institute of Data Science, National Cheng Kung University (NCKU).

• Prof. Li-Wei Kang, Associate Professor, Department of Electrical Engineering, National Taiwan Normal University (NTNU)

• Prof. Chao-Yang Lee, Assistant Professor, Department of Computer Science and Information Engineering, National Yunlin University of Science and Technology

• Prof. Ming-Ching Chang, Associate Professor, Department of Computer Science, College of Engineering and Applied Sciences (CEAS), University at Albany, State University of New York (SUNY).

Challenge Website Link

Kaggle is our competition platform. Please refer to Kaggle: https://www.kaggle.com/competitions/ava-challenge/overview/description

Challenge Description

Video analysis research is important in autonomous driving as it can help with accident prediction. Dashcam videos can provide valuable information for this task. However, there is a lack of well-annotated, large-scale datasets for accident prediction in Asian countries. To address this, a large-scale vehicle collision dataset from Taiwan was collected and labeled for frame- and video-level predictions. The dataset aims to inspire participants from academia and industry to improve safety in autonomous driving. Two tracks of this challenge are vehicle accident prediction and vehicle accident type recognition. More information can be found on the website https://sites.google.com/view/tvcd-tw/home

Besides, all participants are welcome to submit their solutions to ICME workshop. The review is single-blind, and the deadline will be April 28, 2023. 

Paper submission: https://cmt3.research.microsoft.com/User/Login?ReturnUrl=%2FICMEW2023 

Context

According to the research, self-driving cars can only reduce about one-third of major car accidents that occur in the United States out of 5,000 incidents. People typically assume that autonomous driving systems can eliminate human errors while driving. However, many car accidents are still caused by human factors, including perception errors, driver distraction, poor visibility, or slow reactions. Other possible causes are prediction errors, such as misjudging the distance to other vehicles or pedestrians' movements. Some errors may be related to planning and decision-making, such as inappropriate evasive actions or overcompensation when controlling the vehicle. Other accidents can be caused by impairment due to alcohol, drugs, or drowsiness. There are avoidable accidents that are caused by vehicle malfunctions.

Based on the above, we propose a topic for participants to use existing data to predict whether a car accident will occur in the next second to facilitate crisis response in self-driving and improve traffic safety.

The competition host is Advanced Computer Vision Lab (ACVLab), a research team led by Professor Chih-Chung Hsu at National Cheng Kung University. Our lab focuses on introducing machine learning/deep learning into various applications in computer vision and cultivating students with high practical and applied capabilities in deep learning.

Your efforts in this competition could help extend the benefits of road safety. Greater access could further reduce the occurrence of car accidents worldwide.

• Champion:  NVDIA 3090  GPU * 1 for 1st rank only!

• 2023 4/22 Private Leaderboard Score Released                                                                                                                               New!

• 2023 4/21  Note that the Papers Submission Deadline is change to 4/28 !       New!

• 2023 4/11   The  competiton ends in one week , the winner can get one 3090GPU

• 2023 3/3 Join the AVA competition in kaggle: https://www.kaggle.com/competitions/ava-challenge/overview/description