[FoundationposeSampling] No valid value in mask when launching quickstart of isaac_ros_foundationpose 3.2

I want to use foundationpose for object tracking and pose estimation. To understand the way to use foundationpose, I tried quickstart of isaac_ros_foundation.

I successfully completed “Download Quickstart Assets” and “Build (installed binary packages)”. Next, I did “Run Launch File” with rosbag. and “Visualize Results”. And then, I could not see detected objects on View panel like following screenshot.

I checked a terminal for isaac_ros_examples.launch.py. I found the ERROR “ERROR ./gxf/foundationpose/foundationpose_sampling.cpp@310: [FoundationposeSampling] No valid value in mask.” Output of the terminal can be seen from this log file.
foundationpose_log.txt (101.5 KB)

If someone knows the way to cope with this trouble, could you please let me know it? I appreciate any help.

Host computer: Ubuntu 24.04
Container: Ubuntu 22.04 + ROS2 humble

  • this container is built from run_dev.sh. I set CONFIG_IMAGE_KEY=ros2_humble.realsense. And, in build-librealsense.sh, I changed USE_CUDA to false manually because if USE_CUDA is true, RGB frame from RealSense was not shown on Rviz2. But, in case of the trouble above, no realsense device was connected.

I fixed this problem. Just update CUDA version to 12.6.

According to ISAAC ROS 3.2 release note, 3.2 requires CUDA 12.6.

The error “No valid value in mask” in Isaac ROS FoundationPose usually occurs when there is an issue with the mask input or configuration. To fix this, ensure that the mask values are correctly passed in the input data, as missing or incorrect masks can cause this error. Additionally, verify your camera settings, especially if you are using RealSense or another camera for object tracking. Since you updated CUDA to version 12.6, make sure it is compatible with Isaac ROS 3.2, as this version requires CUDA 12.6. Also, double-check the input files, such as rosbag, to ensure they contain valid data. Following these steps should help resolve the issue.