Landmark Detector Regressor Init Workflow

Note

SW Release Applicability: This tutorial is applicable to modules in NVIDIA DRIVE Software releases.

The following code snippet demonstrates the general structure of a program utilizing Landmark Perception to detect landmarks from a single camera. Error handling is left out for clarity.

Initializing the Regressor-based Network

dwMapNetParams mapNetParams{};
dwMapNet_initDefaultParams(&mapNetParams, DW_MAPNET_TYPE_REGRESSOR, sdkHandle)
dwMapNet_initialize(&mapNetHandle, &mapNetParams, sdkHandle);

Modifying Additional Regressor-based Network Initialization Settings and Initialize Detector

uint32_t frameWidth;
uint32_t frameHeight;

// initialize parameters
dwLandmarkDetectorParams params;
dwLandmarkDetector_initializeDefaultParams(&params, frameWidth, frameHeight, sdkHandle));
dwLandmarkDetector_initializeFromMapNet(&landmarkDetectorHandle, mapNetHandle, params, frameWidth, frameHeight, sdkHandle);

// initialize regressor post processing parameters
dwLRNPostProcessingParam regressorParam;
dwLandmarkDetector_initializeDefaultRegressorParams(&regressorParam);
regressorParam.differentClassPenalty = 1.0f;
regressorParam.unitDist = 40.0f;
dwLandmarkDetector_setPostProcessingParam(regressorParam, landmarkDetectorHandle);

Detecting Landmark Markings from a Video Sequence

dwLaneDetection lanes{};
dwLandmarkDetection landmarks{};
while (true)
{
    dwLandmarkDetector_detectLandmarks(&lanes, &landmarks, frame, landmarkDetectorHandle);
}

Releasing the Module Handles when Finished

dwLandmarkDetector_release(landmarkDetectorHandle);
dwMapNet_release(mapNetHandle);

For a more detailed workflow, please refer to the following samples:

• Regressor-based Landmark Detection Sample (MapNet).