Register for the event
Virtual
Webinars & Workshops
Autonomous Vehicles
Exposing Your Data's Blind Spots: Scenario Mining for Safer AV
Aug 27, 2025
9 AM Pacific
Online. Register for the Zoom!
About this event
Stress test AV models by mining real-world edge cases using multimodal data and data-centric tools like FiftyOne.
Host
Autonomous-vehicle perception stacks routinely miss “unknown unknowns” that sit outside their operational-design domain. This webinar shows how to mine those long-tail, safety-critical scenarios hidden inside massive amounts of raw sensor data. We’ll walk through real examples using datasets like Waymo Open, NuScenes, BDD, and KITTI, and help you build data pipelines for training and validating vision models that actually work in the field.
What You Will Learn:
- Why traditional scenario coverage metrics fall short for AV testing
- How to use tools like FiftyOne to visualize data gaps and extract failure-prone samples
- How to mine safety-critical scenarios from large-scale, multimodal datasets
- How to turn natural language descriptions into reusable test cases
- How these workflows reduce development cycles and improve AV system robustness
Who This Is For:
- AV perception, planning, and test engineers working on real-world validation
- Simulation teams looking to scale up coverage
- Machine learning engineers and data scientists focused on model robustness
- Safety and QA engineers interested in scenario-based testing frameworks
- Anyone exploring data-centric approaches to autonomous vehicle development
© 2025 Voxel51 All Rights Reserved