The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 84, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
AdverCPM
Data and models of AdverCPM, a benchmark of benign and adversarial Cooperative Perception Messages (CPMs) for connected vehicles. The code that uses them is GenCPM: it builds the CPMs each vehicle would transmit from its detections, applies the attacks, trains the receiver-side defenses, and evaluates them end to end through late fusion.
Contents
| Folder | What | Used for |
|---|---|---|
raw/ |
CARLA sensor data of every connected vehicle and roadside unit: 4 RGB cameras, LiDAR, pose and ground truth | retraining detectors, new perception models |
frames/pointpillars/ |
per-frame LiDAR detections of every connected vehicle and roadside unit, PointPillars | all CARLA results |
frames/second/ |
the same frames with SECOND detections | the second-perception-model results |
models/defenses/ |
trained temporal models (LSTM, GRU, BiLSTM, CNN-LSTM) per fold and run, with their scalers | inspecting or reusing the models |
models/detectors/cooperscene_pointpillars/ |
PointPillars late-fusion detector trained on CooperScene | detections on CooperScene |
Raw sensor data
raw/Town*/<episode>/<vehicle>.tar (one archive per vehicle, ~2.4 GB) unpacks to
<episode>/<vehicle>/<frame>_camera{0,1,2,3}.png (640x480), <frame>.pcd (LiDAR) and <frame>.yaml (sensor poses and
calibration, ego pose and speed, ground-truth vehicles), in the OPV2V layout read by OpenCOOD. run_info.yaml holds
the scenario settings of the episode. The detections in frames/ were computed from these frames.
Frames
30 CARLA episodes in five towns (Town01, Town02, Town03, Town06: 5 each; Town04: 10), 1,500 ticks of 50 ms each, with
connected vehicles and roadside units spread over a 60 m cluster. Each episode is one archive,
frames/<detector>/Town*/<episode>.tar.gz, which unpacks to
<episode>/run_info.yaml scenario settings (map, number of vehicles, ...)
<episode>/<vehicle>/<frame>_preds.yaml pose of the vehicle, ground-truth vehicles, detected objects
<episode>/<vehicle>/<frame>_score.npy confidence of each detected object (absent when there is no detection)
<frame> is the tick index, written every two ticks (100 ms). The detections come from the PointPillars and SECOND
late-fusion models of the OpenCOOD model zoo, run on each agent separately. The data format is described in
docs/data_format.md.
Use
git clone https://github.com/DhiaNeifar/GenCPM.git && cd GenCPM
conda env create -f environment.yml && conda activate gencpm && pip install -e .
python -m gencpm download --to /path/to/AdverCPM # detections + models; add --parts raw for the sensors
cp configs/paths.example.yaml configs/paths.yaml # then set carla_frames, carla_frames_second, work
bash reproduce/carla.sh all # CPMs, attacks, defenses, late fusion
The CPMs are not stored here: GenCPM generates them from the frames, identically on every machine.
License
MIT.
- Downloads last month
- 314