ImageDatastore from text file

Hi

I need to classify some image data using the trained resnet network.

For the testing data, I need to call them using text file which consist from a sequence directory such as follows:

testingdata.txt:

/DataSet/label1/1.jpg

/DataSet/label1/2.jpg

/DataSet/label1/3.jpg

/DataSet/label1/4.jpg

/DataSet/label2/1.jpg

/DataSet/label2/2.jpg

/DataSet/label2/3.jpg

/DataSet/label2/4.jpg

How to read those image dataset directory from text file, to be used as testing data for deep learning classification.?

Answer 1

[Subhadeep Koley]

Try this code snippet. The below code will read images, whose locations are specified in the testingdata.txt file and will also label them according to their foldernames.

% Read the filenames and put them in cell array
fId = fopen('testingdata.txt');
tline = fgetl(fId);
tlines = cell(0, 1);
while ischar(tline)
    tlines{end+1, 1} = (tline);
    tline = fgetl(fId);
end
% Create an imageDatastore object
imds = imageDatastore(tlines, 'LabelSource', 'foldernames');

[Lin]

thank you very much for your answer.

How about if the directory in the text file list is not uniform such as follows:

testingdata.txt:

/DataSet/label1/imageA/1.jpg

/DataSet/label1/imageB/imageA1/1.jpg

/DataSet/label1/imageC/A/3.jpg

/DataSet/label1/imageD/a/1.jpg

/DataSet/label2/imageA1/1a/2.jpg

/DataSet/label2/imageB/a/2a.jpg

/DataSet/label2/imageC/C1/1c.jpg

/DataSet/label2/imageA/1a/4.jpg

[Subhadeep Koley]

@LS This can be achieved with a bit more processing. There are a lot of ways to achieve this. The below code might help you.

% Read the filenames and put them in cell array
fId = fopen('testingdata.txt');
tline = fgetl(fId);
tlines = cell(0, 1);
while ischar(tline)
    tlines{end+1, 1} = tline;
    tline = fgetl(fId);
end

% Create an imageDatastore object
imds = imageDatastore(tlines);

% Modify image labels of the imageDatastore object
expression = 'label[\d]';
funHandle = @(str)regexpi(str, expression, 'match');
customLabels = cellfun(funHandle, squeeze(imds.Files));
imds.Labels = categorical(customLabels);