Cannot run zed_mono_rect_depth launch file on Jetson AGX Orin Developer Kit due to invalid parameters

I am trying to run yolov8 and zed_mono_rect_depth to both detect objects and measure depth. When I try to include depth, I keep having to reset parameters in the config files that appear to be outdated. I was unable to update the parameter for object_detection.model. The program claims:

[component_container_mt-2] [WARN] [1763679385.686818939] [zed_node]: The value of the parameter ‘object_detection.model’ is not valid: ‘’. Stopping the node

In every config file that includes a model, the value is
model: ‘MULTI_CLASS_BOX_MEDIUM’

This is the current value in the config file called by the launch file, but it is still claiming it is an empty string. It will not let me force a value to it when I run the command. How can I include depth when running the yolo model?

Hello @scruz10,

Welcome to the Isaac ROS forum, and thanks for your post!

Are you planning to use YOLOv8 or the ZED node for object detection? It looks like the zed_node is trying to run its own object detection model, which may conflict with YOLOv8. To avoid this, please disable the ZED object detection by setting object_detection.enable:=false in the ZED parameters (or pass --ros-args -p object_detection.enable:=false when launching). Then, launch the ZED node with depth enabled and run YOLOv8 using the rectified color image as input.

I attempted to set object_detection.enable to false by hard coding it as well as passing an argument to it, but it always keeps it at true regardless of what I do. It appears to be a bug that I cannot work around. This is the command I am running:

ros2 launch isaac_ros_examples isaac_ros_examples.launch.py launch_fragments:=zed_mono_rect_depth,yolov8 model_file_path:=${ISAAC_ROS_WS}/isaac_ros_assets/models/yolov8/yolov8s.onnx engine_file_path:=${ISAAC_ROS_WS}/isaac_ros_assets/models/yolov8/yolov8s.plan interface_specs_file:=${ISAAC_ROS_WS}/isaac_ros_assets/isaac_ros_yolov8/zedx_quickstart_interface_specs.json object_detection.enable:=false