3-D Brain Tumor Segmentation Using Deep Learning - Error using nnet.cnn.TrainingOptionsADAM (line 129)

Question

Moran Artzi 2021-12-6

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1604275-3-d-brain-tumor-segmentation-using-deep-learning-error-using-nnet-cnn-trainingoptionsadam-line-12

评论： vadi su yilmaz 2022-2-15

runing Segment3DBrainTumorUsingDeepLearningExample [on Task01_BrainTumour data]:

options = trainingOptions('adam', ...
    'MaxEpochs',50, ...
    'InitialLearnRate',5e-4, ...
    'LearnRateSchedule','piecewise', ...
    'LearnRateDropPeriod',5, ...
    'LearnRateDropFactor',0.95, ...
    'ValidationData',dsVal, ...
    'ValidationFrequency',400, ...
    'Plots','training-progress', ...
    'Verbose',false, ...
    'MiniBatchSize',miniBatchSize);

I got the following error

Error using nnet.cnn.TrainingOptionsADAM (line 129)
The value of 'ValidationData' is invalid. Invalid transform function defined on datastore.
Error in trainingOptions (line 302)
opts = nnet.cnn.TrainingOptionsADAM(varargin{:});
Caused by:
Error using augmentAndCrop3dPatch
Too many input arguments.

any suggestions?

Thanks a lot

Moran

3 个评论
显示 1更早的评论隐藏 1更早的评论

Moran Artzi 2021-12-6

在 MATLAB Online 中打开

Thanks for your response

miniBatchSize = 8;

dsVal = 
  TransformedDatastore with properties:
      UnderlyingDatastores: {randomPatchExtractionDatastore}
    SupportedOutputFormats: [1×16 string]
                Transforms: {@(patchIn)augmentAndCrop3dPatch(patchIn,outPatchSize,dataSource)}
               IncludeInfo: 0

Lidia Gil Martinez 2022-1-27

Hi,

I have the same issue, is there any solution?

Thanks,

Lidia

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

vadi su yilmaz 2022-2-3

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1604275-3-d-brain-tumor-segmentation-using-deep-learning-error-using-nnet-cnn-trainingoptionsadam-line-12#answer_887650

%%%I had the same problem and solved this with opening the function and change inside of it with code below,

function patchOut = augment3dPatch(patchIn)

flag='validation';

isValidationData = strcmp(flag,'validation');

inpVol = cell(size(patchIn,1),1);

inpResponse = cell(size(patchIn,1),1);

% 5 augmentations: nil,rot90,fliplr,flipud,rot90(fliplr)

fliprot = @(x) rot90(fliplr(x));

augType = {@rot90,@fliplr,@flipud,fliprot};

for id=1:size(patchIn,1)

rndIdx = randi(8,1);

tmpImg = patchIn.InputImage{id};

tmpResp = patchIn.ResponsePixelLabelImage{id};

if rndIdx > 4 || isValidationData

out = tmpImg;

respOut = tmpResp;

else

out = augType{rndIdx}(tmpImg);

respOut = augType{rndIdx}(tmpResp);

end

% Crop the response to to the network's output.

respFinal=respOut(45:end-44,45:end-44,45:end-44,:);

inpVol{id,1}= out;

inpResponse{id,1}=respFinal;

end

patchOut = table(inpVol,inpResponse);

%%% then delete the dsVal from Workspace completely because it will stores the previous one also and create it again then transform with code below,

dsVal = transform(dsVal,@augmentAndCrop3dPatch);

I hope it should be solved

2 个评论
显示无隐藏无

Lidia Gil Martinez 2022-2-7

Hi,

Thanks, it worked, but i get another error:

Error using trainNetwork (line 183)

Invalid transform function defined on datastore.

Caused by:

Error using matlab.io.datastore.TransformedDatastore/read (line 222)

Invalid transform function defined on datastore.

Error using augmentAndCrop3dPatch

Too many input arguments.

vadi su yilmaz 2022-2-15

The problem is the same actually, you should make the same thing for traindata because you should adjust both train data and validation data. There are different way for apply this function to both dataset however I recomend you to adjust the code according to traindata(I adjust and indicate them with bold) and change the name of the function augment3dPatchs (I add s to end you can put another name), paste this function in new script and save,

function patchOut = augment3dPatchs(patchIn)

flag='Training';

traindata = strcmp(flag,'Training');

inpVol = cell(size(patchIn,1),1);

inpResponse = cell(size(patchIn,1),1);

% 5 augmentations: nil,rot90,fliplr,flipud,rot90(fliplr)

fliprot = @(x) rot90(fliplr(x));

augType = {@rot90,@fliplr,@flipud,fliprot};

for id=1:size(patchIn,1)

rndIdx = randi(8,1);

tmpImg = patchIn.InputImage{id};

tmpResp = patchIn.ResponsePixelLabelImage{id};

if rndIdx > 4 || traindata;

out = tmpImg;

respOut = tmpResp;

else

out = augType{rndIdx}(tmpImg);

respOut = augType{rndIdx}(tmpResp);

end

% Crop the response to to the network's output.

respFinal=respOut(45:end-44,45:end-44,45:end-44,:);

inpVol{id,1}= out;

inpResponse{id,1}=respFinal;

end

patchOut = table(inpVol,inpResponse);

%%% then apply

dsTrain = transform(patchds,@augmentAndCrop3dPatchs);

I hope it will solve.

请先登录，再进行评论。

3-D Brain Tumor Segmentation Using Deep Learning - Error using nnet.cnn.TrainingOptionsADAM (line 129)

3 个评论
显示 1更早的评论隐藏 1更早的评论

回答（1 个）

2 个评论
显示无隐藏无

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

3-D Brain Tumor Segmentation Using Deep Learning - Error using nnet.cnn.T​rainingOpt​ionsADAM (line 129)

3 个评论 显示 1更早的评论隐藏 1更早的评论

回答（1 个）

2 个评论 显示 无隐藏 无

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

3-D Brain Tumor Segmentation Using Deep Learning - Error using nnet.cnn.TrainingOptionsADAM (line 129)

3 个评论
显示 1更早的评论隐藏 1更早的评论

2 个评论
显示无隐藏无