shuffle and partition in training, validation and testing of a multi-label combined dataset

Question

Claudio Eutizi 2021-1-15

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/717815-shuffle-and-partition-in-training-validation-and-testing-of-a-multi-label-combined-dataset

回答： Srivardhan Gadila 2021-1-30

在 MATLAB Online 中打开

I want to train a Faster R-CNN with 224x224x3 images that are log-scaled mel spectrograms from the UrbanSound8k dataset

The data is already labeled with rectangles.

here's my code:

%% imds* are M x 2 tables, 1st column 'imageFileName' 2nd 'label_name' with [x,y,height,width] of the rectangular ROIs 
%% Dataset Extraction: Label tables combining
airConditionerLbl = imdsAirConditioner(:,2:end);
carHornLbl = imdsCarHorn(:,2:end);
childrenPlayingLbl = imdsChildrenPlaying(:,2:end);
dogBarkLbl = imdsDogBark(:,2:end);
drillingLbl = imdsDrilling(:,2:end);
engineIdlingLbl = imdsEngineIdling(:,2:end);
gunShotLbl = imdsGunShot(:,2:end);
jackhammerLbl = imdsJackhammer(:,2:end);
sirenLbl = imdsSiren(:,2:end);
streetMusicLbl = imdsStreetMusic(:,2:end);
%% Combining into boxLabelDs
blds = boxLabelDatastore(airConditionerLbl,carHornLbl,childrenPlayingLbl,dogBarkLbl,drillingLbl,engineIdlingLbl,gunShotLbl,jackhammerLbl,sirenLbl,streetMusicLbl);
%% extract imagesFileName from tables
airConditionerFiles = imdsAirConditioner.imageFilename;
carHornFiles = imdsCarHorn.imageFilename;
childrenPlayingFiles = imdsChildrenPlaying.imageFilename;
dogBarkFiles = imdsDogBark.imageFilename;
drillingFiles = imdsDrilling.imageFilename;
engineIdlingFiles = imdsEngineIdling.imageFilename;
gunShotFiles = imdsGunShot.imageFilename;
jackhammerFiles = imdsJackhammer.imageFilename;
sirenFiles = imdsSiren.imageFilename;
streetMusicFiles = imdsStreetMusic.imageFilename;
%% create the imageDatastore
imds = imageDatastore([airConditionerFiles;carHornFiles;childrenPlayingFiles;dogBarkFiles;drillingFiles;engineIdlingFiles;gunShotFiles;jackhammerFiles;sirenFiles;streetMusicFiles]);
%% combine the datastores
cds = combineDatastore(imds,blds);

I want to shuffle and split the data into training, validation and testing and I don't know how and where I can do it in this code.

Can anyone help me?

Thank you.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Claudio Eutizi 2021-1-18

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/717815-shuffle-and-partition-in-training-validation-and-testing-of-a-multi-label-combined-dataset#answer_600200

编辑：Claudio Eutizi 2021-1-18

Can anybody help me with this?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Srivardhan Gadila 2021-1-30

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/717815-shuffle-and-partition-in-training-validation-and-testing-of-a-multi-label-combined-dataset#answer_610283

You can refer to the documentation of the functions partition & shuffle. For the list of other functions supported by the CombinedDatastore object you can refer to Object Functions of the CombinedDatastore.