step 1: Download nuScenes V1.0 full dataset data from HERA on ./data/nuscenes.
step 2: Download nuScenes-lidarseg data from HERA.
step 3: Download (only) the 'gts' from Occ3D-nuScenes
step 4: Create the pkl files:
python tools/create_data_bevdet.py
step 5: Create 2D GT for RenderOcc:
python tools/gen_data/gen_depth_gt.py
python tools/gen_data/gen_seg_gt_from_lidarseg.py
- Download Pretrained model weights
mkdir ckpts
cd ckpts & wget https://github.com/pmj110119/storage/releases/download/v1/bevdet-stbase-4d-stereo-512x1408-cbgs.pth
Folder structure
enderOcc
├── mmdet3d/
├── tools/
├── configs/
├── ckpts/
│ ├── TODO.pth
├── data/
│ ├── nuscenes/
│ │ ├── gts/ # ln -s occupancy gts to this location
│ │ ├── maps/
│ │ ├── samples/
│ │ ├── sweeps/
| | ├── bevdetv2-nuscenes_infos_val.pkl # generated by step 4
| | ├── bevdetv2-nuscenes_infos_train.pkl # generated by step 4
| | ├── seg_gt_lidarseg # generated by step 5
| | ├── depth_gt # generated by step 5
| | ├── lidarseg
| | │ └── v1.0-{mini, test, trainval} <- Contains the .bin files; a .bin file
| | │ contains the labels of the points in a
| | │ point cloud (note that v1.0-test does not
| | │ have any .bin files associated w
| | └── v1.0-{mini, test, trainval}
| | ├── Usual files (e.g. attribute.json, calibrated_sensor.json etc.)
| | ├── lidarseg.json <- contains the mapping of each .bin file to the token
| | └── category.json <- contains the categories of the labels (note that the
category.json from nuScenes v1.0 is overwritten)