Preparing Deep Learning Models for Isaac ROS#

Obtaining a Pre-trained Model from NGC#

The NVIDIA GPU Cloud hosts a catalog of Deep Learning pre-trained models that are available for your development.

  1. Use the Search Bar to find a pre-trained model that you are interested in working with.

  2. Click on the model’s card to view an expanded description, and then click on the File Browser tab along the navigation bar.

  3. Using the File Browser, find a deployable .onnx file for the model you are interested in.

  4. Under the Actions heading, click on the … icon for the file you selected in the previous step, and then click Copy wget command.

  5. Paste the copied command into a terminal to download the model in the current working directory.

Using trtexec to convert an ONNX Model (.onnx) to a TensorRT Plan File (.plan)#

trtexec is already included in the Docker images available as part of the standard Isaac ROS Development Environment. The version of trtexec matches the version of TensorRT installed from the same Docker image.

The per-platform installation paths are described below:

Platform

Installation Path

Symlink Path

x86_64

/usr/src/tensorrt/bin/trtexec

/usr/src/tensorrt/bin/trtexec

Jetson (aarch64)

/usr/src/tensorrt/bin/trtexec

/usr/src/tensorrt/bin/trtexec

Warning

Reading the documentation of trtexec is highly recommended to obtain best performance. In particular, we recommend pay attention to the quantization of the model (e.g. fp32 vs fp16 vs int8).

Converting .onnx to a TensorRT Engine Plan#

Here are some examples for generating the TensorRT engine file using trtexec. In this example, we will use the PeopleSemSegnet Shuffleseg model:

Generate an engine file for the fp16 data type#

mkdir -p /workspaces/isaac_ros-dev/models && \
   /usr/src/tensorrt/bin/trtexec --onnx=/workspaces/isaac_ros-dev/models/peoplesemsegnet_shuffleseg.onnx --saveEngine=/workspaces/isaac_ros-dev/models/peoplesemsegnet_shuffleseg.engine --fp16 --minShapes=input_2:0:1x3x544x960 --optShapes=input_2:0:1x3x544x960 --maxShapes=input_2:0:1x3x544x960

Note

The input ONNX file is specified using the --onnx option. The output file is specified using the --saveEngine option. The specific values used in the command above are retrieved from the PeopleSemSegnet page under the Overview tab. The model input node name can be found in peoplesemsegnet_shuffleseg_cache.txt from File Browser. The tool needs write permission to the output directory.

A detailed explanation of the input parameters is available here.

Generate an engine file for the data type int8#

In this example, we will use the PeopleNet model:

Create the models directory:

mkdir -p /workspaces/isaac_ros-dev/models
Download the calibration cache file:

Note

Check the model’s page on NGC for the latest wget command.

wget --content-disposition 'https://api.ngc.nvidia.com/v2/models/org/nvidia/team/tao/peoplenet/deployable_quantized_onnx_v2.6.3/files?redirect=true&path=resnet34_peoplenet.onnx' -O /workspaces/isaac_ros-dev/models/resnet34_peoplenet.onnx && \
wget --content-disposition 'https://api.ngc.nvidia.com/v2/models/org/nvidia/team/tao/peoplenet/deployable_quantized_onnx_v2.6.3/files?redirect=true&path=resnet34_peoplenet_int8.txt' -O /workspaces/isaac_ros-dev/models/resnet34_peoplenet_int8.txt

/usr/src/tensorrt/bin/trtexec --onnx=/workspaces/isaac_ros-dev/models/resnet34_peoplenet.onnx --saveEngine=/workspaces/isaac_ros-dev/models/resnet34_peoplenet.plan --int8 --calib=/workspaces/isaac_ros-dev/models/resnet34_peoplenet_int8.txt --minShapes=input_1:0:1x3x544x960 --optShapes=input_1:0:1x3x544x960 --maxShapes=input_1:0:1x3x544x960

Note

The calibration cache file (specified using the --calib option) is required to generate the int8 engine file. This file is provided in the File Browser tab of the model’s page on NGC.

Inspecting The Input and Output Binding Names of a Model#

Deep learning models have input_binding_names and output_binding_names. These correspond to the model’s inputs and outputs respectively. These are determined by the model itself during export. To determine this, netron can be used to visualize the ONNX model.

Note

In addition, the TensorRTNode and TritonNode have parameters called input_tensor_names and output_tensor_names, these correspond to the expected tensor names within the ROS 2 TensorList.