Revision history [back]

Hi

We are developing the project which is based on Intel NCS2, OpenVINO and OpenCV. I want to run yolov3 models and OpenCV with NCS2 support to object detection.

I've converted yolov3 models to IR models using the following command:

python3 convert_weights_pb.py --class_names coco.names --data_format NHWC --weights_file yolov3.weights
and
python3 convert_weights_pb.py --class_names coco.names --data_format NHWC --weights_file yolov3-tiny.weights --tiny

python3 /mo_tf.py  --input_model frozen_darknet_yolov3_model.pb --tensorflow_use_custom_operations_config /extensions/front/tf/yolo_v3_tiny.json --input_shape=[1,416,416,3] --data_type FP32 //for CPU
python3 /mo_tf.py  --input_model frozen_darknet_yolov3_model.pb --tensorflow_use_custom_operations_config /extensions/front/tf/yolo_v3_tiny.json --input_shape=[1,416,416,3] --data_type FP16 //for MYRIAD
and
python3 /mo_tf.py  --input_model frozen_darknet_yolov3_model.pb --tensorflow_use_custom_operations_config /extensions/front/tf/yolo_v3.json --input_shape=[1,416,416,3] --data_type FP32 //for CPU
python3 /mo_tf.py  --input_model frozen_darknet_yolov3_model.pb --tensorflow_use_custom_operations_config /extensions/front/tf/yolo_v3.json --input_shape=[1,416,416,3] --data_type FP16 //for MYRIAD

Here is my json file: for yolov3-tiny:

{
    "id": "TFYOLOV3",
    "match_kind": "general",
    "custom_attributes": {
      "classes": 80,
      "anchors": [10, 14, 23, 27, 37, 58, 81, 82, 135, 169, 344, 319],
      "coords": 4,
      "num": 6,
      "mask": [0, 1, 2],
      "jitter": 0.3,
      "ignore_thresh": 0.7,
      "truth_thresh": 1,
      "random": 1,
      "entry_points": ["detector/yolo-v3-tiny/Reshape", "detector/yolo-v3-tiny/Reshape_4"]
    }
  }

for yolov3:

{
    "id": "TFYOLOV3",
    "match_kind": "general",
    "custom_attributes": {
      "classes": 80,
      "coords": 4,
      "num": 9,
      "mask": [0, 1, 2],
      "entry_points": ["detector/yolo-v3/Reshape", "detector/yolo-v3/Reshape_4", "detector/yolo-v3/Reshape_8"]
    }
  }

Process was a complete success

I trying use converted model in OpenCV. Here code:

...
net = readNet(cfg.modelConfiguration, cfg.modelWeights);
net.setPreferableBackend(cfg.backend);
net.setPreferableTarget(cfg.target);
...

//recognize
blob = blobFromImage(frame, cfg.scale, Size(cfg.inpWidth, cfg.inpHeight), Scalar(), false); 
net.setInput(blob);
vector<Mat> outs;
names = getOutputsNames(net);
net.forward(outs, names);

for (size_t i = 0; i < outs.size(); ++i)
{
    Mat prob = outs[i]; //4-dimension Mat    
    //get the coardinates
}

The code was run. But I don't know how to get the coordinates of the bounding.

I tried like this:

    Mat prob = outs[i];
    Mat detectionMat(prob.size[2], prob.size[3], CV_32F, prob.ptr<float>());

            float* data = (float*)detectionMat.data;
            for (int j = 0; j < detectionMat.rows; ++j, data += detectionMat.cols)
            {  
                Mat scores = detectionMat.row(j).colRange(5, detectionMat.cols);
                Point classIdPoint;
                double confidence;
                // Get the value and location of the maximum score
                minMaxLoc(scores, 0, &confidence, 0, &classIdPoint);
                if (confidence > cfg.confThreshold)
                {
                    ObjResult result;
                    cout << "confidence " << confidence << endl;
                    cout << "classIdPoint.x " << classIdPoint.x << endl;
                    // cout << "classes[classId] " << classes[classIdPoint.x] << endl;

                    result.centerX = (int)(data[0] * frame.cols);
                    result.centerY = (int)(data[1] * frame.rows);
                    result.width = (int)(data[2] * frame.cols);
                    result.height = (int)(data[3] * frame.rows);
                    result.confidence = confidence;
                    result.classId = classIdPoint.x;
                    result.className = classes[classIdPoint.x];
                    ret.objResults.push_back(result);
                }
            }

or:

    Mat prob = outs[i];
    Mat detectionMat(prob.size[2], prob.size[3], CV_32F, prob.ptr<float>());
            for(int c=0; c<numlabels; c++)
            {
                int labelnum = (size_t)(detectionMat.at<float>(c, 1))-1;
                labelnum = (labelnum < 0) ? 0 : (labelnum > numlabels) ? numlabels : labelnum;
                cout << "labelnum: " << labelnum << endl;

                float confidence = detectionMat.at<float>(c, 2);
                cout << "confidence: " << confidence << endl;
                if(confidence > 0.5) 
                {
                    int classId = static_cast<int>(detectionMat.at<float>(c, 1));
                    int left = static_cast<int>(detectionMat.at<float>(c, 3) * frame.cols);
                    int top = static_cast<int>(detectionMat.at<float>(c, 4) * frame.rows);
                    int right = static_cast<int>(detectionMat.at<float>(c, 5) * frame.cols);
                    int bottom = static_cast<int>(detectionMat.at<float>(c, 6) * frame.rows);

                    ObjResult result;
                        result.width = right - left;
                        result.height = top - bottom;
                        result.centerX = left + result.width/2;
                        result.centerY = bottom + result.height/2;
                        result.confidence = confidence;
                        result.classId = classId;
                        result.className = classes[classId];
                        ret.objResults.push_back(result);

                }
            }

But it's not proper way.

How can i get the coordinates of the bounding? Have you any sample code?