Simliar to OpenPCDet
repo, 3DTrans
repo also provides a quick sequence demo to show the groundtruth or prediction results of on the custom point cloud data to visualize the continuous results.
You need to follow the INSTALL.md to install the 3DTrans
repo successfully.
- Make sure you have already installed the
Open3D
. If not, you could install it as follows:
pip install open3d==0.15.2
- Run the demo with your custom point cloud data and groundturth infos as follows:
python show_squence_demo/demo.py --data_path ${POINT_CLOUD_DATA} --seq_id ${SEQUENCE_ID}
--func once or nuscenes
Here ${SEQUENCE_ID}
is the sequence id of the data, such as "000076" of ONCE, or "n015-2018-07-18-11-07-57+0800" of nuScenes.
Here ${POINT_CLOUD_DATA}
can be in any of the following format:
ONCE:
├──once_data
│ │── XXX(squence_id)
│ │ │── XXX(squence_id).json
│ │ │── lidar_roof
│ │ │ │── *.bin
nuScenes:
├──nuscenes_data
│ ├── nuscenes_infos_10sweeps_train.pkl
│ ├── samples
│ │ ├── LIDAR_TOP
│ │ │ │── *.pcd.bin
Then you could see the groundtruth results with visualized point cloud as follows: