converters for the generic 4D radar dataset which is here.
- convert the rosbag to a generic folder of a 4D radar dataset sequence
python3 folder_to_bag.py
- convert the content of a generic folder of a 4D radar dataset sequence to a rosbag
python3 bag_to_folder.py
The snail_radar_SDK.py script enables visualization and synchronization of point clouds and stereo images. With this script, you can:
- View multiple point clouds (e.g., ars548/points, eagleg7/enhanced, eagleg7/pcl, xt32) in separate Open3D windows.
- Synchronized left and right images from ZED2i stereo cameras in OpenCV windows.
We recommend creating a virtual env with miniconda.
conda create -n test_env python=3.10
conda activate test_env
pip install numpy==1.24.4 opencv-python open3d
sudo apt-get install xdotool
To use the snail_radar_SDK.py script for visualization, run: python3 snail_radar_SDK.py <folder_path>
- <folder_path>: Path to a sequence of the dataset.
Example:
python3 snail_radar_SDK.py /path/to/dataset/20230920/data1
This will:
- Display point clouds from the specified sequence in multiple Open3D windows.
- Show the left and right ZED2i stereo images in synchronized OpenCV windows.
The frame definitions are given here. The frames used in sensor messages are listed in table. The extrinsic calibrations are provided in the matlab scritps folder, and the ROS1 roslaunch files with name tf_static.launch in each sequence's folder.
The intrinsic parameters of the zed2i stereo camera are in the topics of each bag, /zed2i/zed_node/left_raw/camera_info and /zed2i/zed_node/right_raw/camera_info.
The python scripts require a ros1 distribution to run, e.g., ros1 noetic which can be installed for ubuntu <=20.04 following instructions.
For higher ubuntu versions say 22.04, a docker container or the robostack can be used to run this program. Both are pretty easy to install and use.
The matlab scripts require a MATLAB >=2007a installation.