How to deploy to Isaac ROS image_segmentation with TAO5?

Hello,

We can success to deploy to Isaac ROS image_segmentation (GitHub - NVIDIA-ISAAC-ROS/isaac_ros_image_segmentation: Hardware-accelerated, deep learned semantic image segmentation) using below exported “etlt file” with TAO4.
GitHub - NVIDIA-AI-IOT/robot_freespace_seg_Isaac_TAO: In this workflow we demonstrate using SDG + TAO for a freespace segmentation application

But we cannot export “etlt file” latest TAO5 by this script.

Question:
Could you tell us recommended way to deploy to IsaacROS image_segmentation with TAO5 environment such as below notebook example?


Best regards,

Hello,

Are there any update on this?

We would appreciate if you tell us how to generate xx.etlt from xx.onnx by exported TAO5?


Best regards

Hi @mat-h and sorry for the late reply,

Have you tried the latest Isaac ROS 2.0 release? We made significant changes to the packages and documentation.

New documentation: NVIDIA Isaac ROS — isaac_ros_docs documentation
Image segmentation documentation: Isaac ROS Image Segmentation — isaac_ros_docs documentation

Let me know if the new release everything is fixed.

Raffaello

Hi,

Thank you very much for your kindness.
I really appreciate your help.

I checked new github documentation like below;
https://nvidia-isaac-ros.github.io/repositories_and_packages/isaac_ros_image_segmentation/isaac_ros_unet/index.html#quickstart

But it seems to be same guide like “tao-converter with .etlt”.
Are there any clue to use .onnx from TAO5 export model file?

Best regards,

Hi,

It seems to be success to generate TensotRT engine from onnx file by “trtexec” command.
Is it proper way on Isaac ROS ?

Best regards,