Download RADIal raw dataset data HERE. Prepare the KITTI format RADIal data by running
Download data and convert format
cd EchoFusion
# download RADIal data to ./data/radial
python tools/radial/convert_radial.py --root-path ./data/radial --out-dir data/radial_kitti_format
Copy camera_calib.npy
Copy camera_calib.npy (shared by all frames) to the folder of the KITTI format RADIal data by running:
mkdir data/radial_kitti_format/calibs
cp RADIal/DBReader/examples/camera_calib.npy data/radial_kitti_format/calibs
Generate info pickle files
python tools/radial/create_data.py radial
Copy CalibrationTable.npy used for data processing
cp RADIal/SignalProcessing/CalibrationTable.npy data/radial_kitti_format
Prepare 3D labels
Download 3D labels of RADIAl from HERE and save the txt files to data/radial_kitti_format/labels_x/
. Then create info pickle files for 3D labels using:
python tools/radialx/create_data.py radialx
Folder structure
EchoFusion
├── projects/
├── tools/
├── configs/
├── ckpts/
│ ├── resnet50_bev.pth
├── data/
│ ├── radial/
│ │ ├── RECORD@2020-11-21_11.54.31/
│ │ ├── RECORD@2020-11-21_11.58.53/
│ │ ├── ...
│ │ ├── RECORD@2020-11-22_12.56.17/
| | ├── CalibrationTable.npy
| | ├── labels.csv
│ ├── radial_kitti_format/
│ │ ├── calibs/
│ │ ├── images/
│ │ ├── ImageSets/
│ │ ├── labels/
│ │ ├── labels_x/
| | ├── lidars/
| | ├── radars_adc/
| | ├── radars_pcd/
| | ├── radars_ra/
| | ├── radars_rd/
| | ├── radars_rt/
| | ├── CalibrationTable.npy
| | ├── radial_infos_test.pkl
| | ├── radial_infos_train.pkl
| | ├── radial_infos_trainval.pkl
| | ├── radial_infos_val.pkl
| | ├── radialx_infos_test.pkl
| | ├── radialx_infos_train.pkl
| | ├── radialx_infos_trainval.pkl
| | ├── radialx_infos_val.pkl
Download and unzip data
Download the first 20 sequences of KRadar dataset from HERE and unzip.
cd EchoFusion
# download KRadar data to ./data/k-radar-zip/1-20
# unzip data to ./data/k-radar
python tools/kradar/unzip_data.py
Prepare sparse radar tensor
Use our modified KRadar codebase to prepare sparse radar tensor. It is required to change LIST_DIR
and DIR_SPARSE_CB
to your own path to k-radar.
cd K-Radar-main
python datasets/kradar_detection_v1_0.py
cd ..
mkdir data/k-radar/ImageSets
python K-Radar-main/tools/train_test_splitter/train_val_splitter.py
python K-Radar-main/tools/train_test_splitter/train_test_splitter.py
Generate info pickle files
python tools/kradar/create_data.py kradar
Folder structure
EchoFusion
├── projects/
├── tools/
├── configs/
├── ckpts/
│ ├── resnet50_bev.pth
├── data/
│ ├── radial/
│ ├── radial_kitti_format/
│ ├── k-radar/
│ │ ├── 1/
│ │ │ ├── cam-front/
│ │ │ ├── cam-left/
│ │ │ ├── cam-rear/
│ │ │ ├── cam-right/
│ │ │ ├── images/
│ │ │ ├── info_calib/
│ │ │ ├── info_label/
│ │ │ ├── os1-128/
│ │ │ ├── os2-64/
│ │ │ ├── radar_dense_pcd/
│ │ │ ├── radar_ra/
│ │ │ ├── radar_zyx_cube/
│ │ │ ├── sp_rdr_cube/
│ │ │ ├── time_info/
│ │ │ ├── description.txt
│ │ ├── 2/
│ │ ├── ...
│ │ ├── 19/
│ │ ├── 20/
| | ├── ImageSets/
| | ├── kradar_infos_test.pkl
| | ├── kradar_infos_train.pkl
| | ├── kradar_infos_trainval.pkl
| | ├── kradar_infos_val.pkl