What is the most efficient way to create a large image database from MATLAB plots or figures for use in Deep Learning?

I am working with a large dataset and have created a looped code that:

• Creates a colourmap / surface plot from Continuous Wavelet Transform (cwt) data.
• Alters the image size and axes to make them suitable for use in 'Deep Learning Toolbox'.
• Writes the image, as a .png, to a file name and location using imwrite.
• Repeats for each set of data.

I have approximately 340,000 images to create and as you can imagine, this has been taking >2days to create and even though each file size is ~30kb, the total folder size is ~10Gb.

The code is as shown:

(Missing: DB2data = raw data; file_list = file name list; save_path = directory save path)

for m = 1:length(DB2data)  
    %%      SIGNAL PROCESSING
      
    fig = figure('visible','off');
    [cfs,frq]=cwt(DB2data(m,:),fs);
    surf(t,frq,abs(cfs),'edgecolor','none')
    colormap parula    %Set colour to 'parula'
    view(2)
    
    %%      IMAGE SETTINGS
    
    ylim([0.5 40])  %Set frequency axis
    zlim([0 56.5])   %Set colourbar limits
    xlim([0 5]) %Set time axis
    axis off    %Remove axes
    colorbar off    %Remove colorbar
    set(gca,'LooseInset',get(gca,'TightInset'));    %Remove background
    set(gca,'units','normalized','position',[0 0 1 1]);     %Remove additional background
    fig2 = getframe(fig);   %Get 'image' from figure
    image = fig2.cdata;     %Imgage data
    image = imresize(image, [227 227]); %Resize image
    imwrite(image, fullfile(save_path, strcat(file_list(m),'.png')));   %Save image
    close(fig)  %Close figure

end 

Does anyone have any suggests on a more efficient way to create a large image database like this?

Thanks in advance.

Answer 1

[Srivardhan Gadila]

The following suggestions might help you:

1. Use parfor instead of the regular for loop, this would save time.
2. While loading the generated dataset, make use of the imageDatastore or imageSet

[Cameron Huggins]

Thank you for the suggestions. I will give this a go and get back to you!