Training the RCNN Detector with multivariable from groundtruth

Hi guys

Hope you are doing well

I have one code which is for Traning the RCNN object detector for detecting the smoke

i made dataset for labelling set of images which have many variable such as sky cars people smoke however i want in my code that to train the detector to detect smoke and not to detect the other variables which are labelled

currently in my code i can only make selecetlabels for smoke and traning the detector with smoke but i want to know how to select other variables and train the detector not to detect or select other variables

herein my code

load('gTruth.mat')
smokedetection = selectLabels(gTruth,'smoke');
if isfolder(fullfile('TrainingData'))
    cd TrainingData
else
    mkdir TrainingData
end 
addpath('TrainingData');
options = trainingOptions('sgdm', ...
  'MiniBatchSize', 32, ...
  'InitialLearnRate', 1e-6, ...
  'MaxEpochs', 10);
layers = [
    imageInputLayer([32 32 3],"Name","imageinput")
    convolution2dLayer([5 5],32,"Name","conv","BiasLearnRateFactor",2,"Padding",[2 2 2 2],"WeightsInitializer","narrow-normal")
    maxPooling2dLayer([3 3],"Name","maxpool","Stride",[2 2])
    reluLayer("Name","relu")
    averagePooling2dLayer([3 3],"Name","avgpool","Stride",[2 2])
    fullyConnectedLayer(2,"Name","fc_rcnn","BiasL2Factor",1,"BiasLearnRateFactor",10,"WeightLearnRateFactor",20,"WeightsInitializer","narrow-normal")
    softmaxLayer("Name","softmax")
    classificationLayer("Name","classoutput")];
trainingData = objectDetectorTrainingData(smokedetection,'SamplingFactor',1,...
'WriteLocation','TrainingData');
detector = trainRCNNObjectDetector(trainingData, layers, options,  ...
        'NegativeOverlapRange', [0 0.3]);
save('Detector.mat','detector');

Answer 1

[Kritika Bansal]

Hi,

As per my understanding, you need a way to train the RCNN object detector to detect classes other than smoke.

Assuming you are using MATLAB R2019a, you can use the following functions to select the ground truth data for a set of labels:

1. selectLabelsByGroup(): https://www.mathworks.com/help/vision/ref/groundtruth.selectlabelsbygroup.html
2. selectLabelsByName(): https://www.mathworks.com/help/vision/ref/groundtruth.selectlabelsbyname.html