Tao-converter issues

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

I am trying to use TAO toolbox to train a PointPillars model. I have gone through the tao-tutorial, tao_tutorials/notebooks/tao_launcher_starter_kit/pointpillars/specs/pointpillars.yaml at main · NVIDIA/tao_tutorials · GitHub . Now I have a tlt file, but I need to convert that to use in the ros node provided here. GitHub - NVIDIA-AI-IOT/ros2_tao_pointpillars: ROS2 node for 3D object detection using TAO-PointPillars. . I want to convert my model, but I am unclear on the options I should use. It looks like the converter requires dimensions, but I don’t see anywhere in the notebook or the yaml file that details the dimensions, and outputs.

I am using TensorRT 8.6 in the tao docker files, as well as the TensorRT 24.05 tag on docker.

Thanks

3

Please export .tlt file to .onnx file and then use trtexec to generate engine. Refer to Incorrect pointpillar inference results - #6 by Morganh.

4

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

5

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