DriveWorks SDK Reference
4.0.0 Release
For Test and Development only

DNN Plugin Sample


The DNN Plugin sample loads and runs an MNIST network with a custom implementation of the max pooling layer.

PoolPlugin.cpp implements this layer and defines the functions that are required by DriveWorks to load a plugin. This file is then compiled as a shared library, which is then loaded by sample_dnn_plugin executable at runtime.

Running the Sample

The command line for the sample is:


Accepts the following optional parameters:

./sample_dnn_tensor --tensorRT_model=[path/to/TensorRT/model]


        Specifies the path to the NVIDIA<sup>&reg;</sup> TensorRT<sup>&trade;</sup>
        model file.
        The loaded network is expected to have a output blob named "prob".
        Default value: path/to/data/samples/dnn/<gpu-architecture>/mnist.bin, where <gpu-architecture> can be `volta-discrete` or `volta-integrated` or `turing`.
This sample loads mnist digit recognition network from:
  • data/samples/dnn/volta if the detected GPU has VOLTA architecture.
  • data/samples/dnn/turing if the detected GPU has TURING architecture.


The sample creates a window, displays a white screen where a hand-drawn digit will be recognized via aforementioned MNIST network model.

Digit being recognized using DNN with a plugin

Additional Information

For more information, see DNN Plugins.