Skip to content

Latest commit

 

History

History
28 lines (23 loc) · 1.63 KB

DATASET_PREPARATION.md

File metadata and controls

28 lines (23 loc) · 1.63 KB
Raw

Dataset Preparation

Baraja Spectrum-Scan™ Dataset

Please download the Baraja Spectrum-Scan™ Dataset and organise the downloaded files as follows:

SEE-VCN
├── data
│   ├── baraja
│   │   │── ImageSets
│   │   │── test
│   │   │   ├──pcd & masks & image & calib & label
│   │   │── infos
│   │   │   ├──baraja_infos_test.pkl
...

KITTI, Waymo, nuScenes

For KITTI, Waymo, nuScenes datasets, we follow closely to the data preparation of OpenPCDet.

For training of our networks, we use a subset of the nuScenes and Waymo datasets.

  • nuScenes. We sort the dataset by number of cars in each scene and choose the top 100 scenes. This give us 4025 frames for training.
  • Waymo. We use every 30th frame to reduce the training data size. This gives us 5267 frames.

For testing on nuScenes and KITTI, we use the original validation set.

VC-ShapeNet

You don't need this dataset for SEE-VCN as we provide the pretrained models, however if you'd like to download our VC-ShapeNet dataset, you can download it here. You can also download our kitti/nuscenes/waymo test dataset here.