Download the following files and unzip under the kitti_depth
folder:
data_depth_annotated, data_depth_velodyne, data_depth_selection
Finally, download kitti raw images by:
cd datasets/kitti_depth
wget https://github.com/youmi-zym/CompletionFormer/files/12575038/kitti_archives_to_download.txt
wget -i kitti_archives_to_download.txt -P kitti_raw/
cd kitti_raw
unzip "*.zip"
The overall data directory is structured as follows:
kitti_depth
├──data_depth_annotated
| ├── train
| └── val
├── data_depth_velodyne
| ├── train
| └── val
├── data_depth_selection
| ├── test_depth_completion_anonymous
| |── test_depth_prediction_anonymous
| └── val_selection_cropped
└── kitti_raw
├── 2011_09_26
├── 2011_09_28
├── 2011_09_29
├── 2011_09_30
└── 2011_10_03
Go to benchmark_data_generator/relative_depth
. Run render_kitti_image.py
or render_kitti_flipud.py
(remember to replace the PATH_TO_KITTI
variable
to the directory you downloaded the KITTI dataset into).