DriveWorks SDK Reference
3.5.78 Release
For Test and Development only

ParkNet Workflow

Initialization

ParkNet DNN module initialization starts with initializing ParkNet-DNN parameters structure dwParkNetParams by providing the current DriveWorks context. Such initialization generally sets the current platform parameters such as GPU operating precision.

The rest of the parameters, such as the specific ParkNet model to be used (if different by default) is set by the user.

ParkNet-DNN initialization takes a ParkNet-DNN module handle and ParkNet-DNN parameter structure dwParkNetParams as input. To initialize NVIDIA ParkNet module the following call is used:

The ParkNet-DNN module can be initialized with a custom-version of a parking space perception DNN. The following code snippet exemplifies such initialization of a ParkNet module with a custom model that is optimized to perform inference on GPU in FP16-precision:

dwParkNetParams parkNetParams{};
dwParkNet_initDefaultParams(&parkNetParams, ctx);
parkNetParams.networkModel = DW_PARKNET_MODEL_CUSTOM;
parkNetParams.networkPrecision = DW_PRECISION_FP16;
parkNetParams.networkCustomData = "path/to/my/parknet/model";
CHECK_DW_ERROR(dwParkNet_initialize(&parkNet, &parkNetParams, context));
Note
dwParkNet_initDefaultParams() assumes that "path/to/my/parknet/model" file is already present in data/resources sub-directory.

Query Network Specs

To receive the input blob dimension that the initialized ParkNet expects:

dwStatus dwParkNet_getInputBlobsize(
dwBlobSize* inputBlobsize,

To retrieve the output blob dimension:

To retrieve the meta-data containing preprocessing configuration for the network:

ParkNet-DNN Module Reset

To reset the ParkNet-DNN module:

ParkNet Module Release

To release the resources associated with the ParkNet-DNN module:

For more detailed workflow, please refer to the parking space detection sample: Parking Space Detection Sample .