Since the dataset is generated by a simulation, you are free to use it under the conditions of the MIT license.
The full dataset is 1.3 TB in size. Most of this is the LiDAR pointclouds. You can download the dataset via https as described below.
The dataset can be downloaded in parts, each split into a train and a test archive.
SHA-256 checksums for all archives are available in SHA256SUMS.txt.
| Part | Description | Train | Test |
|---|---|---|---|
| Base | Front Camera + additional sensors (GNSS, IMU, Weather) + Anomaly Labels | download (137G) | download (86G) |
| Camera Extended | Left, Right, Rear camera sensor, including instance segmentation masks and anomaly-labels | download (427G) | download (267G) |
| LiDAR | Pointclouds with anomaly and semantic segmentation labels | download (227G) | download (138G) |
| KITTI | LiDAR labels in KITTI format | download (22G) | download (12G) |
| Depth Maps | Depth maps for all cameras | download (74G) | download (48G) |
You can download and set up the base dataset with
cd my/data/dir/
wget https://data.carlanomaly.de/v1/carlanomaly-base-train.tar.gz
tar -xzf carlanomaly-base-train.tar.gz
Afterwards, download additional data with
cd my/data/dir/
wget https://data.carlanomaly.de/v1/carlanomaly-<name>-<split>.tar.gz
tar -xzf carlanomaly-<name>-<split>.tar.gz