datasets

Used Datasets

In general we provide support for rosbags and the kitti dataset. For each dataset we assume the following hierarchical structure: dataset_name/<path_to_rosbag> or for KITTI its original sturcture dataset_name/sequence/scan. Here, sequences are numbered according to 00, 01, ...99. After prepocessing, scans will be numbered according to 00000...99999. An example for preprocessing a rosbag can be seen with the DARPA SubT dataset, the KITTI example can be seen in the KITTI secion.

Rosbag - DARPA SubT Dataset Example

Download the DARPA SubT Rosbags: link

mkdir $PWD/datasets/darpa/
# Link taken from https://bitbucket.org/subtchallenge/subt_reference_datasets/src/master/
wget https://subt-data.s3.amazonaws.com/SubT_Urban_Ckt/a_lvl_1.bag -O $PWD/datasets/darpa/00.bag

Structure

Run preprocessing

Pull the rosbag at the above link, and put it to <delora_ws>/datasets/darpa/<name>.bag. Rename it to <delora_ws>/datasets/darpa/00.bag (or 01...99.bag if you have multiple sequences). In the file ./config/deployment_options.yaml set datasets: ["darpa"]. Preprocessing can then be run with the following command:

preprocess_data.py

If your files are placed somewhere else, simply adapt the path in ./config/config_datasets.yaml (global or local w.r.t. to python working directory).

KITTI Dataset

LiDAR Scans

Download the "velodyne laster data" from the official KITTI odometry evaluation ( 80GB): link. Put it to <delora_ws>/datasets/kitti, where kitti contains /data_odometry_velodyne/dataset/sequences/00..21.

Groundtruth poses

Please also download the groundtruth poses here. Make sure that the files are located at <delora_ws>/datasets/kitti, where kitti contains /data_odometry_poses/dataset/poses/00..10.txt.

Run preprocessing

In the file ./config/deployment_options.yaml set datasets: ["kitti"]. Then run

preprocess_data.py

Custom Dataset

Just follow the above procedure for custom datasets. Any sequence of rosbags can be used.

Visualize Processed Dataset

The point cloud and its estimated normals for a dataset can be visualized using the following command:

visualize_pointcloud_normals.py

With this command, the first 100 scans with its normals are published under the topics /lidar/points and /lidar/normals in the frame lidar and can be visualized in RVIZ.