Problem with RCNN Detector

1 次查看（过去 30 天）

Abdussalam Elhanashi 2019-8-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/477261-problem-with-rcnn-detector

评论： Dinesh Yadav 2019-8-26

Hi Guys

I have a problem with RCNN detector as i am trying to train the detector for smoke images which is basiclly with layers Herein the code :-

any solution

Thank you in advance

load('gTruth.mat')
load('layers')
smokedetection = selectLabels(gTruth,'smokealarm');
 if isfolder(fullfile('TrainingData'))
    cd TrainingData
else
    mkdir TrainingData
end 
addpath('TrainingData');
trainingData = objectDetectorTrainingData(smokedetection,'SamplingFactor',4,...
'WriteLocation','TrainingData');
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")
    convolution2dLayer([5 5],32,"Name","conv_1","BiasLearnRateFactor",2,"Padding",[2 2 2 2],"WeightsInitializer","narrow-normal")
    reluLayer("Name","relu_1")
    averagePooling2dLayer([3 3],"Name","avgpool","Stride",[2 2])
    convolution2dLayer([5 5],64,"Name","conv_2","BiasLearnRateFactor",2,"Padding",[2 2 2 2],"WeightsInitializer","narrow-normal")
    reluLayer("Name","relu_2")
    averagePooling2dLayer([3 3],"Name","avgpool_1","Stride",[2 2])
    fullyConnectedLayer(64,"Name","fc","BiasLearnRateFactor",2,"WeightsInitializer","narrow-normal")
    reluLayer("Name","relu_3")
    fullyConnectedLayer(2,"Name","fc_rcnn","BiasL2Factor",1,"BiasLearnRateFactor",10,"WeightLearnRateFactor",20,"WeightsInitializer","narrow-normal")
    softmaxLayer("Name","softmax")
    classificationLayer("Name","classoutput")];
options = trainingOptions('sgdm', ...
  'MiniBatchSize', 32, ...
  'InitialLearnRate', 1e-6, ...
  'MaxEpochs', 10);
imds=trainingData
detector = trainRCNNObjectDetector(trainingData, layers, options, 'NegativeOverlapRange', [0 0.3]);
save('Detector.mat','detector');

and the errors are :-

Error using trainNetwork (line 165)
The class names of layer 13 must match the class names of the training data. The training data class names are given by
categories(Y), where Y are the training data labels.
Error in rcnnObjectDetector.train (line 234)
            [net, info] = trainNetwork(dispatcher, layers, opts);
Error in trainRCNNObjectDetector (line 278)
[detector, ~, info] = rcnnObjectDetector.train(trainingData, lgraphOrLayers, options, params);
Error in TrainingsmokedetectionwithRCNN (line 22)
detector = trainRCNNObjectDetector(trainingData, layers, options, 'NegativeOverlapRange', [0 0.3]);

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Dinesh Yadav 2019-8-26

"The class names of layer 13 must match the class names of the training data. The training data class names are given by categories(Y), where Y are the training data labels". The error you displayed is self explanatory. Kindly output and compare the training data class labels with categories(Y) word to word i.e. Spacing, Capitals etc.

请先登录，再进行评论。

请先登录，再回答此问题。