MuSoHu: Multi-Modal Social Human Navigation Dataset.

Dependencies :

ROS Noetic.

Python 3.8.10.

LiDAR driver Velodyne VLP-16.

Stereo camera driver ZED2.

Create a workspace and clone sources

mkdir -p catkin_ws/src; cd catkin_ws/src; catkin_init_workspace
git clone https://github.com/ros-drivers/velodyne.git
git clone --recursive https://github.com/stereolabs/zed-ros-wrapper.git
git clone https://github.com/stereolabs/zed-ros-interfaces.git

Install dependencies with rosdep :

cd catkin_ws; rosdep install --from-paths . --ignore-src --rosdistro=noetic

Build and source

cd catkin_ws; catkin_make; source devel/setup.bash

Launch visualization:

In terminal 1 :

roslaunch musohu_package musohu_suite.launch

Record data:

In terminal 2 :

python3 record.py

Link to TF Tree / URDF

Downloading the data

Thanks to Jiaxu, you can download the dataset using this notebook.

Parsing bag files

To parse bag files and create samples please follow this guide. To load the data, here is a PyTorch dataloader example which loads ego-centeric images.