nexar
Request a demo
< Back
AI

Nexar’s Open Dataset and Crash Prediction Challenge: Pushing the Boundaries of AI-Driven Road Safety

Orly Zvitia |
Nexar’s Open Dataset and Crash Prediction Challenge: Pushing the Boundaries of AI-Driven Road Safety

The Nexar Crash Prediction Challenge is live! Using 1,500 real-world driving videos, we’re inviting the global AI community to predict collisions before they happen. Help us push the boundaries of AI for safer, smarter mobility.

At Nexar, we turn cars into vision sensors to better understand the physical world. Our service uses videos from dashcams shared by everyday drivers to collect real-time information about roads and traffic. This helps make driving safer, improves city planning, and supports new transportation technologies that use AI. 

Last week, we announced an exciting new collaboration with NVIDIA to accelerate AI-powered innovation in autonomous driving. By integrating Nexar's crowdsourced vision data with NVIDIA's cutting-edge AI computing platforms, we are paving the way for next-generation autonomous driving models to enhance vehicle safety, decision-making, and overall road intelligence.

But there’s more. As part of our commitment to AI Innovation, we are also engaging with the broader AI research community in multiple ways: collaborating with top AI and AV researchers, participating in key AI conferences, and recently releasing an open dataset of real-world collision and near-collision videos. On top of that, we’re proud to sponsor the Nexar Crash Prediction Challenge, which leverages the open dataset and is an effort to push the boundaries of automotive collision prediction. 

The Nexar Open Dataset: Fueling AI Research

At the core of this initiative is our Open Dataset, a carefully curated collection of 1,500 real-world driving videos, including actual collisions, near-collisions, and normal driving videos

Designed to drive advancements in predictive modeling and road safety applications, this dataset is hosted on Hugging Face and available under a permissive license. This open-access approach empowers researchers, engineers, and AI innovators to explore real-world driving scenarios in ways never before possible. Much like our partnership with NVIDIA, which helps scale AI-powered mobility solutions, this dataset bridges the gap between research and industry applications, driving real-world impact in autonomous systems, ADAS, and road safety AI.

The Nexar Dashcam Crash Prediction Challenge

To push the boundaries of AI-driven collision detection, we launched a global Kaggle competition, challenging participants to develop predictive models to determine whether a collision is imminent based on a video sequence.

We provide an in-depth look into the dataset structure, collection, processing and the reasoning behind key decision choices in the Nexar Dashcam Collision Prediction Dataset and Challenge paper.

If you are participating in the competition, this paper will help you better understand the dataset and refine your approach!

We’re now halfway into this competition; a triple-digit number of teams have signed up and are busily submitting models whose performance keeps increasing. This is going to be a nail-biter!

The competition runs until May 4, 2025, offering AI researchers, data scientists, and mobility innovators the opportunity to develop, test, and showcase their models on real-world driving data.

What’s Next?

Once the Nexar Crash Prediction Challenge concludes, Nexar will evaluate the top-performing models and announce the winner. The first-place prize includes an all-expenses-paid trip to Portugal and a VIP visit to our Porto offices, where the winning team will present their project and key findings to Nexar’s leadership and R&D team.

Beyond the competition, Nexar will share the challenge’s key insights across multiple platforms. We will present the results at CVPR, one of the world’s leading AI and computer vision conferences, publish an academic paper, and release a widely distributed blog post recognizing the top three winners and their contributions to advancing collision prediction.

How to Get Involved

This is a unique opportunity to apply AI to real-world driving data and make a tangible impact on road safety. If you’re working in AI, computer vision, or autonomous systems, don’t miss out—join the challenge today!