top of page

Access 500 Million km of Real World Data to verify the robustness of your algorithms

Screenshot 2024-10-18 at 15.03.20.png

Human drivers are remarkably good at avoiding accidents — even when others make mistakes.

Resembler captures and models these human-level avoidance behaviors to help you measure how your AV or ADAS system performs using real-world data.

Through our unique A–B verification tests, you can directly compare your algorithm’s responses against real human data in identical traffic scenarios.

Human danger recognition is universal — people everywhere instinctively sense and avoid risk.

Resembler models this behavior using Singapore as its living testbed — a uniquely dense, multi-modal city where every traffic scenario converges, from urban and industrial zones to ports, airports, and highways — providing the perfect foundation to validate safety in any city worldwide.

Wavy Abstract Background

Products

1

Search
 

Find and analyze real-world accidents and incidents to build your own verification tests.


We provide track-based models that can be imported directly into OpenSCENARIO and CARLA.
Each event includes:

​​​​​​

  • full environment and actor trajectories,
     

  • human perception cues showing when danger is detected,
     

  • comparisons to similar real-world cases.

2

Upload

Upload your own accident and automatically convert it to OpenSCENARIO format.


Discover similar cases in our database and explore accidentology-based synthetic cues to vary severity and validate the robustness of your algorithms.

​

​

​

​

​

​

3

Synthetic Verification (coming soon)

Use our real-world database to verify your synthetic accident datasets and ensure statistical and behavioral realism in simulation-based testing.

​

​

​​

​

​

​

​

​

​

​

​

Our Numbers

5

Years of Data

km Coverage

569,197,895

Europe, US and Asia

3

bottom of page