Skip to content

Latest commit

 

History

History
44 lines (37 loc) · 1.54 KB

File metadata and controls

44 lines (37 loc) · 1.54 KB

Data Preparation

  1. Download nuScenes dataset to ./data/nuscenes.

  2. Download Info Files. Unzip the data_nusc.tar.gz and move the files into ./data/nuscenes.

  3. Download occupancy labels 'gts' from Occ3D-nuScenes.

  4. Download checkpoints and save in ./ckpts.

  5. Download pretrained LLM and save in ./ckpts.

Folder structure

SparseOccVLA
├── projects/
|   ├── configs
├── mmdetection3d/
├── tools/
├── ckpts/
|   ├── InternVL2-2B
|   ├── cascade_mask_rcnn_r50_fpn_coco-20e_20e_nuim_20201009_124951-40963960.pth
|   ├── stage1_3d_600q_lidar.pth
|   ├── stage1_4d_600q.pth
|   ├── stage3_4d_600q_forecasting.pth
├── data/
│   ├── nuscenes/
│   │   ├── maps/
│   │   ├── samples/
│   │   ├── sweeps/
│   │   ├── gts/
│   │   ├── v1.0-test/
│   │   ├── v1.0-trainval/
│   │   ├── conv/
│   │   ├── desc/
│   │   ├── keywords/
│   │   ├── vqa/
│   │   ├── nuscenes2d_ego_temporal_infos_train.pkl
│   │   ├── nuscenes2d_ego_temporal_infos_val.pkl
│   │   ├── data_dict_sample.pkl
│   │   ├── data_dict_subset_B.json
│   │   ├── data_dict_subset_B_val.pkl