MATLAB Answers

Error using Semantic Segmentation Using Deep Learning

18 views (last 30 days)
Vasilis Giannoglou
Vasilis Giannoglou on 10 Dec 2019
Answered: Vasilis Giannoglou on 16 Dec 2019
I was trying to replicate the example in Semantic Segmentation Using Deep Learning
Error using matlab.io.datastore.PixelLabelDatastore.label2cat (line 1030)
Pixel label image has RGB-triplet pixel label IDs instead of scalar pixel label IDs.
At first I got it at
pximds = pixelLabelImageDatastore(imdsTrain,pxdsTrain, ...
'DataAugmentation',augmenter);
But then I didn't include the augmenter and there was no problem there. Afterwards, the same again in:
[net, info] = trainNetwork(pximds,lgraph,options);
I searched it in google, it seems other people also had the same, but there was no answer to it. It's just the example proposed in mathworks

  0 Comments

Sign in to comment.

Answers (2)

Sourav Bairagya
Sourav Bairagya on 13 Dec 2019
I have tried to replicate that same example (mentioned in the provided link) on MATLAB R2019b and it sucessfully ran.
On MATLAB 2019a, "deeplabv3plusLayers" function is not available. Here I have ran the same example using U-Net and it sucessfully ran.
% Create U-Net
lgraph = unetLayers(imageSize, numClasses);
imageFreq = tbl.PixelCount ./ tbl.ImagePixelCount;
classWeights = median(imageFreq) ./ imageFreq;
pxLayer = pixelClassificationLayer('Name','labels','Classes',tbl.Name,'ClassWeights',classWeights);
lgraph = replaceLayer(lgraph,"Segmentation-Layer",pxLayer);
The above mentioned errors have not been produced in any of the cases.

  0 Comments

Sign in to comment.


Vasilis Giannoglou
Vasilis Giannoglou on 16 Dec 2019
If I copy-paste all the commands in a seperate .m file, I get the above errors. If I open the
openExample('deeplearning_shared/SemanticSegmentationUsingDeepLearningExample')
then it runs ok. I really don't know why

  0 Comments

Sign in to comment.

Products


Release

R2019a