Freespace segmentation on Orin NX 8GB

Robotics & Edge Computing Robotics - Isaac Isaac ROS
1

Hi,

I am interested in using GitHub - NVIDIA-ISAAC-ROS/isaac_ros_freespace_segmentation: NVIDIA-accelerated, deep-learned freespace segmentation on my Jetson Orin NX 8GB (1 DLA). My goal is to create an occupancy grid for nav2.

However, I see that it depends on Bi3D and that Bi3D only lists the Orin NX 16GB version in system requirements.

  • Will I be able to run it on my 8GB version?
  • Is there any workaround to still use freespace segmentation like using ESS ESS DNN Stereo Disparity | NVIDIA NGC or does freespace segmentation only run with Bi3D.
  • Is there any isaac ros module then to be able to perform obstacle avoidance // provide occupancy grid to nav2 using ESS?
  • What would be the best way to proceed knowing that I must use the Orin NX 8GB?

Thank you

2

Hi @eloi3

Let me reply question by question.

  • For an optimal result, we suggest using a 16 GB version, but it can also work on your Orin NX 8Gb version. The performance summary for all devices is available on this page Performance Summary — isaac_ros_docs documentation
  • You can combine the Isaac ROS free space segmentation and Isaac ROS DNN stereo disparity package. If I misunderstood your question, please let me know.
  • You can use Isaac ROS vslam and Isaac ROS nvblox to perform the obstacle avoidance. Please take a look at this demo with Isaac Sim that perform with Nav2 Isaac Sim Examples — isaac_ros_docs documentation
  • My suggestion is to start by following our tutorials step by step and learning the capability of each Isaac ROS package. Then, start designing your own ROS 2 launch file, combining the packages you need.

A list of packages that can be interesting for you:

Best,
Raffaello

3

Thank you for the information I will review the packages

4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.