How can I obtain matrix of image in detectnet_camera.cpp?

Hi everyone. Can somebody tell me what is the meaning of “(float*)imgRGBA” in the line of “if( net->Detect((float*)imgRGBA, camera->GetWidth(), camera->GetHeight(), bbCPU, &numBoundingBoxes, confCPU))” in the detectnet_camera.cpp?
I am using detectnet by Jetson Inference to do face detection. However, the given facenet model is not good enough for some scene. I want to know the meaning of “(float*)imgRGBA” so that I can inference the image by other models. I am a beginner of CUDA programming and openCV, I have tried transforming the (float*)imgRGBA to cvMat but failed. Can somebody tell me how can I get the matrix of image in detectnet_camera.cpp (like (float*)imgRGBA)?


1. Here is the document for function detectNet::Detect():

2. Jetson_inference can change model with the argument input directly:

3. If you want to save an image for debugging, check here for example: