isaac_ros_bi3d_freespace

Source code on GitHub.

Quickstart

1. Set up your development environment by following the instructions here.

2. Clone isaac_ros_common and this repository under ${ISAAC_ROS_WS}/src.

   cd ${ISAAC_ROS_WS}/src
   

   git clone https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_common.git
   

   git clone https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_freespace_segmentation.git
   

3. Pull down a rosbag of sample data:

   git clone https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_depth_segmentation.git
   

   cd ${ISAAC_ROS_WS}/src/isaac_ros_depth_segmentation &&
   git lfs pull -X "" -I "resources/rosbags/bi3dnode_rosbag"
   

4. Launch the Docker container using the run_dev.sh script:

   cd ${ISAAC_ROS_WS}/src/isaac_ros_common &&
   ./scripts/run_dev.sh
   

5. Install this package’s dependencies.

sudo apt-get install -y ros-humble-isaac-ros-bi3d-freespace

6. Download model files for Bi3D (refer to the Model Preparation section for more information):

   mkdir -p /tmp/models/bi3d &&
   cd /tmp/models/bi3d &&
   wget 'https://api.ngc.nvidia.com/v2/models/nvidia/isaac/bi3d_proximity_segmentation/versions/2.0.0/files/featnet.onnx' &&
   wget 'https://api.ngc.nvidia.com/v2/models/nvidia/isaac/bi3d_proximity_segmentation/versions/2.0.0/files/segnet.onnx'
   

7. Convert the .onnx model files to TensorRT engine plan files (refer to the Model Preparation section for more information):

   If using Jetson (Generate engine plans with DLA support enabled):

   /usr/src/tensorrt/bin/trtexec --saveEngine=/tmp/models/bi3d/bi3dnet_featnet.plan \
   --onnx=/tmp/models/bi3d/featnet.onnx \
   --int8 --useDLACore=0 --allowGPUFallback &&
   /usr/src/tensorrt/bin/trtexec --saveEngine=/tmp/models/bi3d/bi3dnet_segnet.plan \
   --onnx=/tmp/models/bi3d/segnet.onnx \
   --int8 --useDLACore=0 --allowGPUFallback
   

   If using x86_64:

   /usr/src/tensorrt/bin/trtexec --saveEngine=/tmp/models/bi3d/bi3dnet_featnet.plan \
   --onnx=/tmp/models/bi3d/featnet.onnx --int8 &&
   /usr/src/tensorrt/bin/trtexec --saveEngine=/tmp/models/bi3d/bi3dnet_segnet.plan \
   --onnx=/tmp/models/bi3d/segnet.onnx --int8
   

   Note

   The engine plans generated using the x86_64 commands will also work on Jetson, but performance will be reduced.

8. Run the launch file to spin up a demo of this package:

   cd /workspaces/isaac_ros-dev
   

   ros2 launch isaac_ros_bi3d_freespace isaac_ros_bi3d_freespace.launch.py featnet_engine_file_path:=/tmp/models/bi3d/bi3dnet_featnet.plan \
   segnet_engine_file_path:=/tmp/models/bi3d/bi3dnet_segnet.plan \
   max_disparity_values:=10
   

9. Open a second terminal inside the Docker container:

   cd ${ISAAC_ROS_WS}/src/isaac_ros_common && \
     ./scripts/run_dev.sh
   

10. Visualize the occupancy grid in RViz.

    Start RViz:

    rviz2
    

    In the left pane, change Fixed Frame to base_link.

    In the left pane, click the Add button, then select By topic followed by Map to add the occupancy grid.

Try More Examples

To continue your exploration, check out the following suggested examples:

• Tutorial with RealSense
• Tutorial with Zed
• Tutorial with Isaac Sim

API

Usage

ros2 launch isaac_ros_bi3d_freespace isaac_ros_freespace_segmentation.launch.py base_link_frame:=<"name of base link"> camera_frame:=<"name of camera frame"> f_x:=<"focal length in pixels in x dimension"> f_y:=<"focal length in pixels in y dimension"> grid_width:=<"desired grid width"> grid_height:=<"desired grid height"> grid_resolution:=<"desired grid resolution">

ROS Parameters

ROS Parameter	Type	Default	Description
base_link_frame	std::string	base_link	The name of the tf2 frame attached to the robot base
camera_frame	std::string	camera	The name of the tf2 frame attached to the camera
f_x	double	0.0	The focal length in pixels in x dimension
f_y	double	0.0	The focal length in pixels in y dimension
grid_width	int	100	The width of the output occupancy grid, in number of cells
grid_height	int	100	The height of the output occupancy grid, in number of cells
grid_resolution	double	0.01	The resolution of the output occupancy grid, in meters per cell

ROS Topics Subscribed

ROS Topic	Interface	Description
freespace_segmentation/mask_in	stereo_msgs/DisparityImage	The input disparity image, with pixels corresponding to ground labeled as 0

Limitation: For all input images, both the height and width must be an even number of pixels.

ROS Topics Published

ROS Topic	Interface	Description
freespace_segmentation/occupancy_grid	nav_msgs/OccupancyGrid	The output occupancy grid, with cells marked as 0 if free