Main Content

loadHRNETObjectKeypointDetector

Load HRNet object keypoint detector model for code generation

Since R2023b

Description

net = loadHRNETObjectKeypointDetector(filename) loads a pretrained HRNet object keypoint detector model net from the MAT file filename. filename must be a compile-time constant.

Note

This functionality requires Deep Learning Toolbox™ and the Computer Vision Toolbox™ Model for Object Keypoint Detection. You can download and install the Computer Vision Toolbox Model for Object Keypoint Detection from Add-On Explorer. For more information about installing add-ons, see Get and Manage Add-Ons.

example

Examples

collapse all

Use the loadHRNETObjectDetector function to load an HRNet object detector and generate C code for the network. You must have a pretrained hrnetObjectKeypointDetector object saved as a variable named detector in the file trainedHRNetObjectKeypointDetector.mat.

Load the pretrained hrnetObjectKeypointDetector object from the MAT file.

load("trainedHRNetObjectKeypointDetector.mat");

Create an entry-point function myHRNetObject.m that accepts an image and the bounding boxes of image objects as input. The entry-point function performs these operations:

  • Defines a persistent variable called myDetector. The persistent variable prevents reconstructing and reloading the network object during subsequent calls to the myHRNetObject function.

  • Loads the detector from the file trainedHRNetObjectKeypointDetector.mat into the myDetector variable using the loadHRNETObjectDetector function.

  • Detects object keypoints, confidence scores, and validity of the objects in the input test image imTest using the detector.

function [keypoints,scores,valid] = myHRNetObject(imTest,bboxes)
  persistent myDetector;
    if isempty(myDetector)
        myDetector = loadHRNETObjectKeypointDetector("trainedHRNetObjectKeypointDetector.mat");
    end
  [keypoints,scores,valid] = detect(myDetector,imTest,bboxes);
end

Read a test image from a graphics file.

imTest = imread("visionTeam.jpg");

Specify the bounding boxes of objects in the test image.

bboxes = [28.999092	48.454803	136.47928	364.54520;
157.66484	34.789597	101.99452	375.76184;
261.84634	39.663254	123.77393	345.26978;
388.49371	49.597092	111.15561	328.06561;
506.54700	51.744110	140.14929	351.37390;
656.18372	49.514694	134.45197	363.48529];

Create a coder.config (MATLAB Coder) configuration object for MEX code generation and set the target language to C++. On the configuration object, set the DeepLearningConfig property to the target library "mkldnn", for MKL-DNN targets. For more information about the supported target libraries for C++ code generation in deep learning networks, see Generate Generic C/C++ Code for Deep Learning Networks (MATLAB Coder).

Note

The HRNet object keypoint detector deep learning network architecture contains the spaceToDepthLayer. Code generation for spaceToDepthLayer is not supported for target library "none".

The codegen (MATLAB Coder) function must determine the size, class, and complexity of MATLAB® function inputs. Use the -args option to specify the input test image imTest and object bounding boxes bboxes to the entry-point function. Use the -config option to pass the code configuration object myHRNetObject to codegen.

cfg = coder.config("mex");
cfg.TargetLang = "C++";
cfg.DeepLearningConfig = coder.DeepLearningConfig(TargetLibrary = "mkldnn");
codegen -args {imTest,bboxes} -config cfg myHRNetObject -report;

The codegen command places all the generated files in the codegen folder. The folder contains the C code for the entry-point function myHRNetObject.c, header and source files containing the C class definitions for the network, weight, and bias files.

Compute the object keypoint locations, keypoint confidence scores, and keypoint validity in the test image by calling the code-generated MEX file myHRNetObject_mex.

[objKeypoints,keypointScores,keypointValid] = myHRNetObject_mex(imTest,bboxes)
objKeypoints(:,:,1) =

   80.0250   88.0543
   84.5691   83.3077
   70.6329   83.3077
   93.9612   88.0543
   56.8991   92.8010
  113.2518  121.2811
   48.4184  130.7745
  137.5927  173.4946
   68.3165  192.4813
   95.8852  178.2413
  123.7577  178.2413
  115.7833  239.9482
   69.3291  239.9482
  140.9344  330.1351
   75.9997  334.8818
  151.8454  406.0820
   91.4549  406.0820


objKeypoints(:,:,2) =

  208.9805   60.3294
  218.4503   55.4367
  204.0877   55.4367
  228.1306   60.3294
  194.6179   60.3294
  248.2275  104.3640
  176.4147  104.3640
  253.9619  148.3986
  172.6793  153.2913
  201.1940  143.5059
  234.9171  153.2913
  236.0744  207.1114
  193.0919  212.0041
  237.9681  295.1806
  194.9856  300.0733
  239.6515  373.4643
  191.7762  373.4643


objKeypoints(:,:,3) =

  332.8377   58.8094
  337.1445   54.3137
  323.9408   54.3137
  346.1359   63.3051
  315.2329   58.8094
  360.1899  103.7664
  298.3838   94.7750
  374.4328  153.2192
  282.0071  153.2192
  362.1739  198.1762
  303.4464  126.2449
  353.6549  211.6633
  309.5482  207.1676
  355.2610  288.0902
  298.0451  288.0902
  352.1825  351.0300
  290.6599  355.5257


objKeypoints(:,:,4) =

  449.0307   76.9031
  453.1255   72.6314
  440.4873   68.3597
  461.5805   76.9031
  432.2976   76.9031
  479.1093  115.3483
  416.3604  115.3483
  488.5369  166.6085
  404.8717  166.6085
  489.4212  209.3254
  401.5726  209.3254
  472.7766  213.5971
  426.7607  213.5971
  474.1913  281.9441
  423.9922  281.9441
  471.2460  341.7477
  421.1353  346.0194


objKeypoints(:,:,5) =

  573.6482   89.4887
  582.4121   80.3383
  569.0730   84.9135
  591.4658   84.9135
  560.2125   89.4887
  614.7282  126.0901
  543.0713  126.0901
  633.8019  180.9923
  526.2200  176.4171
  625.9077  231.3193
  522.7077  222.1689
  599.1330  235.8945
  549.8689  235.8945
  600.5824  304.5222
  551.2217  299.9470
  601.9352  368.5747
  552.3812  354.8492


objKeypoints(:,:,6) =

  720.0562   83.7954
  729.2193   79.0625
  715.3233   79.0625
  743.3170   88.5283
  706.2611   88.5283
  762.8539  135.8571
  698.0060  135.8571
  773.5303  202.1174
  680.5878  187.9188
  760.7441  254.1791
  677.0656  239.9805
  742.1152  249.4462
  700.3264  244.7133
  743.9312  334.6381
  706.6735  325.1723
  731.5486  405.6313
  712.8188  396.1655


keypointScores =

    0.9671    0.9648    0.9650    0.9617    0.9565    0.9564
    0.9809    0.9810    0.9627    0.9769    0.9846    0.9839
    0.9762    0.9953    0.9727    0.9730    0.9699    0.9814
    0.9646    0.9943    0.9740    0.9795    0.9659    0.9850
    0.9305    0.9845    0.9636    0.9675    0.9577    0.9677
    0.9603    0.9220    0.9320    0.9558    0.9397    0.8990
    0.9044    0.9620    0.8772    0.9620    0.9502    0.9549
    0.9584    0.9620    0.9544    0.9133    0.9080    0.9309
    0.9316    0.9707    0.9516    0.9333    0.9222    0.9300
    0.7920    0.9553    0.9716    0.9492    0.9421    0.9558
    0.9104    0.7432    0.9154    0.9533    0.9711    0.9333
    0.8087    0.8730    0.8530    0.8723    0.8607    0.8483
    0.7811    0.8598    0.8277    0.8443    0.8671    0.8340
    0.8606    0.9081    0.9165    0.9512    0.8997    0.9069
    0.8365    0.9004    0.9179    0.9260    0.9611    0.8890
    0.8982    0.9391    0.9052    0.9220    0.9360    0.9669
    0.7941    0.9696    0.8999    0.9185    0.9210    0.9494


keypointValid =

  17×6 logical array

   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1
   1   1   1   1   1   1 

Input Arguments

collapse all

Filename of the MAT file containing the pretrained HRNet object keypoint detector, specified as a character vector or string scalar. The MAT file must exist on the MATLAB path and contain only the network to be loaded.

This input argument must be a compile-time constant.

Data Types: char | string

Output Arguments

collapse all

HRNet object keypoint detector network, returned as a hrnetObjectKeypointDetector object.

Version History

Introduced in R2023b