CarlAnomaly

Supplementary Resources

Dev Kit

Code for training and evaluating your model on the benchmark is provided in this GitHub repository.

CarlaGen

We also release the code that was used to simulate the scenarios - "CarlaGen". This is basically a CARLA-centered framework that combines the hydra configuration framework with the CARLA Python API. It defines some callbacks which you can use to cause events or capture data from various sensors that you just describe in yaml configuration files.

This allows you to, for example, create a new version of the dataset with higher-resolution camera images or additional sensors, such as dynamic vision sensor (DVS), optical flow, or radar.

Baseline Models

Labeling Interface

The interface used to label the dataset.