head-pose-estimation-adas-0001 use in opecv dnn?

[fti-sbargule]

I try to head-pose-estimation-adas-0001 this model use in cv::dnn::net

std::string config_pose = "model/pose/head-pose-estimation-adas-0001.xml";
std::string model_pose = "model/pose/head-pose-estimation-adas-0001.bin";
net_pose = cv::dnn::readNet(config_pose, model_pose);
 
inputblob_pose = cv::dnn::blobFromImage(image, scalefactor_pose, cv::Size(width_pose, height_pose), meanVal, false, false);
cv::Mat net_pose.setInput(inputblob_pose);
cv::Mat result_pose = net_pose.forward();
std::cout << "Pose detection size -: " << result_pose.size() << std::endl;

Output :-

Pose detection size -:  [1 x 1]

• Output is one float value.
• How to find angle of yaw, roll, pitch.

[Wovchena]

I'm not familiar with cv::dnn API, but I can see that forward() takes a string as arg where you can specify a name for layer which output is needed to get. There's also CV_WRAP void forward(OutputArrayOfArrays outputBlobs, const String& outputName = String()); which fills outputBlobs with all model outputs