Error when using semanticseg on a imagedatastore

This is the error information:

error using DAGNetwork/calculateActivations (line 23)

The input images for activations must be of size equal to or greater than [32 32 1].

error DAGNetwork/activationsSeries (line 239)

Y = this.calculateActivations(X, layerID, 1, varargin{:});

error SeriesNetwork/activations (line 673)

Y = this.UnderlyingDAGNetwork.activationsSeries(X, layerID, varargin{:});

error semanticseg>iClassifyImagePixels (line 441)

allScores = activations(net, X, params.PixelLayerID, ...

error semanticseg>iProcessImageDatastoreInParallel (line 713)

L = iClassifyImagePixels(X, net, params);

error semanticseg (line 209)

filenames = iProcessImageDatastoreInParallel(imds, net, params);

The code:

testdatastore=imageDatastore('parpath\');
%montage(testdatastore)
%Try to segment the test image.
tic
pxds = semanticseg(testdatastore,net,'ExecutionEnvironment',"cpu")
%C = semanticseg(Img,net,'ExecutionEnvironment',"cpu");
%B = labeloverlay(testImage,C);
%imshow(B)
pxdsC=="activated";
X=ans;
Num=size(regionprops(X,"BoundingBox"))
imshow(X)
t2=toc

However, when I use semanticseg processing single picture, everything goes well and gives a result in several seconds. The code is almost the same:

C = semanticseg(Img,net,'ExecutionEnvironment',"cpu");

The image is about 4000*4000 JPG. When I use semanticseg on a imagedatastore of 32*32 or 64*64 images, there are pixellabel outcomes, but the outcomes are totally different from when I am not using parallel computing and definitely they are wrong. I also dont know what's wrong with it.

[Srivardhan Gadila]

Can you copy some images(2 or 3 on which "C = semanticseg(Img,net,'ExecutionEnvironment',"cpu");" is working) to a new folder and try runnig the code on images present in the new folder using the imageDatastore.