Skip to content

Latest commit

 

History

History
81 lines (73 loc) · 2.58 KB

data.md

File metadata and controls

81 lines (73 loc) · 2.58 KB

Dataset Preparation

NuScenes

1. Download nuScenes V1.0 full dataset data HERE. Folder structure:

DAOcc
├── data/
│   ├── nuscenes/
│   │   ├── maps/
│   │   ├── samples/
│   │   ├── sweeps/
│   │   ├── v1.0-test/ (optional)
│   │   ├── v1.0-trainval/

2. Occ3D-nuScenes

For Occupancy Prediction task, download (only) the gts.tar.gz from Occ3D-nuScenes and arrange the folder as:

DAOcc
├── data/
│   ├── nuscenes/
│   │   ├── gts/ (new)
│   │   ├── maps/
│   │   ├── samples/
│   │   ├── sweeps/
│   │   ├── v1.0-test/
│   │   ├── v1.0-trainval/

3. (Optional) SurroundOcc

Download the generated dense occupancy labels (resolution 200x200x16 with voxel size 0.5m) from SurroundOcc, then unzip and place them in the data folder:

DAOcc
├── data/
│   ├── nuscenes/
│   │   ├── gts/
│   │   ├── maps/
│   │   ├── samples/
│   │   ├── sweeps/
│   │   ├── v1.0-test/
│   │   ├── v1.0-trainval/
│   ├── nuscenes_occ/ (new)

4. (Optional) OpenOccupancy

Download the v0.1 annotation package from OpenOccupancy, then unzip and place them in the data folder:

DAOcc
├── data/
│   ├── nuscenes/
│   │   ├── gts/
│   │   ├── maps/
│   │   ├── samples/
│   │   ├── sweeps/
│   │   ├── v1.0-test/
│   │   ├── v1.0-trainval/
│   ├── nuscenes_occ/ (optional)
│   ├── nuScenes-Occupancy/ (new)

5. Download the generated train/val pickle files and put them in data. Folder structure:

DAOcc
├── data/
│   ├── nuscenes/
│   │   ├── gts/ (new)
│   │   ├── maps/
│   │   ├── samples/
│   │   ├── sweeps/
│   │   ├── v1.0-test/ (optional)
│   │   ├── v1.0-trainval/
│   │   ├── nuscenes_infos_train_w_3occ.pkl (new)
│   │   ├── nuscenes_infos_val_w_3occ.pkl (new)
│   ├── nuscenes_occ/ (optional)
│   ├── nuScenes-Occupancy/ (optional)

Waymo

TBD