Dataset Preparation
January 12, 2026 · View on GitHub
Download Dataset
DSEC-3DOD
- DSEC-3DOD dataset can be downloaded from the hugging face.
DSEC (Train Event)
Download the DSEC Train split from the official website:
- DSEC download page: https://dsec.ifi.uzh.ch/dsec-datasets/download/
Required data
Only the following is needed:
- Split: train
- Sequence:
zurich_city - Modalities:
- events
- disparity
The dataset should have the following format:
data
├── dsec
│ ├── detection
│ │ ├── zurich_city_00_a_00
│ │ ├── zurich_city_00_a_01
│ │ ├── ...
│ ├── detection_train_sample_new.txt
│ ├── detection_val_sample_new.txt
│ ├── train
│ │ ├── zurich_city_00_a
│ │ │ │ ├── events
│ │ │ │ │ ├── left
│ │ │ │ │ │ ├── events.h5
│ │ │ │ │ │ └── rectify_map.h5
│ │ │ │ │ └── right
│ │ │ │ │ ├── events.h5
│ │ │ │ │ └── rectify_map.h5
│ │ │ │ └── disparity
│ │ │ │ ├── event
│ │ │ │ └── timestamps.txt
│ │ ├── zurich_city_00_b
│ │ └── ...
Stereo Event Voxel
python data/dsec_raw_event_parsing.py --dataset_path data/dsec/train
python data/dsec_left_right_voxel_npz.py
Preprocessing (KITTI-like format)
- Generate the data infos by running the following command:
python -m liga.datasets.dsec.lidar_dsec_dataset create_dsec_infos