It appears imageLabeler cannot handle a transformed datastore? - matlab

Question

The Matlab app imageLabeler is supposed to support the following format:
imageLabeler(imgStore)
I have an imgStore, defined as follows:
imds = imageDatastore(cellArrayOfImageFilenames);
imgStore = transform(imds, #(x)demosaic(x,'rggb'));
I have to do this, because my images are stored as bayer encoded images, and this is the only way I've figured out to get the imgStore to return these images as 3 channel RGB images. However, when I try and initalize imageLabeler, I get this error:
>> imageLabeler(imgStore)
Error using imageLabelerInternal
Expected input name to be one of these types:
char
Instead its type was matlab.io.datastore.TransformedDatastore.
Error in vision.internal.imageLabeler.imageLabelerInternal
Error in imageLabeler (line 58)
vision.internal.imageLabeler.imageLabelerInternal(varargin{:});
TLDR:
How do I get imageLabeler to handle my bayer encoded images?

Answer 1

The way to fix this, is with the imageDatastore 'ReadFcn' parameter. The documentation for imageDatastore explicity tells you to NOT do this, as it slows down Neural Network stuff. Here's the Matlab doc text:
Using ReadFcn to transform or pre-process 2-D images is not
recommended. For file formats recognized by imformats, specifying
ReadFcn slows down the performance of imageDatastore. For more
efficient ways to transform and pre-process images, see Preprocess
Images for Deep Learning (Deep Learning Toolbox).
So, all that said, here's the workaround:
imgStore = imageDatastore(cellArrayOfImageFilenames ...
, 'ReadFcn', #(x)demosaic(imread(x),'rggb')));