SplitEachLabel Error for divide Images into Train and Test Data for Faster R-CNN

I have 297 Grayscale images (Drone) and I would Like Divide Them into 3 parts (train-test and validation) for Object Detection Using Faster R-CNN method and I used SplitEachLabel Command in Matlab2018a but after write below codes:

imds = imageDatastore('C:\Users\hosein\My Documents\MATLAB\DroneImages');

[trainds,testds,valds] = splitEachLabel(imds,0.6,0.2,'randomized');

matlab showed error: "Input folders or files contain non-standard file extensions.Use FileExtensions Name-Value pair to include the non-standard file extensions."

and when I wrote this code:

[trainds,testds,valds] = splitEachLabel(imds,0.6,0.2)

matlab showed error:"splitEachLabel requires non-empty Labels property.."

[Julian Lopez Baasch]

Be sure you loaded the data properly. You can check that by computing the data size. Try with

>> size(imds)

If the output is of the shape 0xsomething, then you hadn't load the data properly. Maybe your path is wrong.

[Michael Mashevsky]

Did you ever find an anwwer to this question?

Answer 1

[Obaid Rafiq Jan]

The error is quite straightforward; 'splitEachLabel' requires a label from the 'imageDatastore' object which is not defined in your code. One of the input arguments for 'imageDatastore' is its 'LabelSource', by default it is set to none and you can specify any label if you want, or you can simply let the label definition be 'foldernames' which takes the name of your imageDatastore folder and provides that as the label. You can try:

dataDir = fullfile('C:\Users\hosein\My Documents\MATLAB');
imDir = fullfile(dataDir, 'DroneImages');

imds = imageDatastore(imDir, 'LabelSource', 'foldernames');
[trainds,testds,valds] = splitEachLabel(imds,0.6,0.2);