Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.

TROUBLE with face-person-detection-retail-0002

liu__zhiqi
Beginner
549 Views

Hi, 

     I am using face-person-detection SSD model, trying to identify both people and faces in security scenarios, when the cameras are typically installed high above. I use the simple image as input, and this model was able to identify faces, but NOT standing people. There are indeed output returned with label == 1, which means this is a person; however, the probability is ~0.1. Why?

 

 

0 Kudos
4 Replies
liu__zhiqi
Beginner
549 Views

 

output from this model is: 

The net outputs blob with shape: [1, 1, N, 7], where N is the number of detected bounding boxes. For each detection, the description has the format: [image_id, label, conf, x_min, y_min, x_max, y_max]

  • image_id - ID of the image in the batch
  • label - predicted class ID
  • conf - confidence for the predicted class
  • (x_min, y_min) - coordinates of the top left bounding box corner
  • (x_max, y_max) - coordinates of the bottom right bounding box corner.

 

I was able to detect faces with label == 2. just not people. Thank you for helping out.

0 Kudos
Monique_J_Intel
Employee
549 Views

Hi Zhiqi,

Did you resolve your issue by looking in the in-package documentation /opt/intel/computer_vision_sdk/deployment_tools/documentation/docs/face-person-detection-retail-0002.html?

Also, as you can see the pose coverage is standing upright, parallel to image plane so If the camera is to high the accuracy will be poor.

Kind Regards,

Monique Jones

0 Kudos
liu__zhiqi
Beginner
549 Views

Hi, 

My image is taken at the horizontal position. I found the issue but I don't know how to solve it: the face-person-detection-retail-0002 has two output layers, including detection_out_face and detection_out_pedestrain. So in code, when I use outputInfo.begin()->first, it gives me the layer for face detection. I don't know how to enumerate to the pedestrian output later with outputInfo. I tried the end(), or [], either worked. 

Can you please let me know if a model contains several SSD, how can I get through all output layers?

 

0 Kudos
liu__zhiqi
Beginner
549 Views

OK, this is the way to fix it. thanks.

 for (auto& output : outputInfo)
 {
  auto outputName = output.first;
  auto& output_ptr = output.second;

 

 

0 Kudos
Reply