Error using trainNetwork VALUESET must be the same class as DATA.

Hi. I'm trying to train a U-Net and the error that i'm getting is:

Error using trainNetwork

VALUESET must be the same class as DATA.

Caused by:

Error using matlab.io.datastore.ImageDatastore/read

Error using ReadFcn @customReadFcn for file:

C:\Users\New User\OneDrive - ump.edu.my\Desktop\NewLabelOriginal\c0002s0002t01_lesionsmooth_115.mat

Error using categorical

VALUESET must be the same class as DATA.

Error in customReadFcn (line 7)

data = categorical(labels, [0, 1], {'Background', 'Tumor'});

I have attached the codes in this question.

Could someone lend a hand in troubleshooting this error?ubleshooting this error? Thank you for your time.

[Walter Roberson]

We need c0002s0002t01_lesionsmooth_115.mat for testing -- as outside observers we cannot assume that c0002s0002t01_lesionsmooth_115.mat has the same class() of data as c0001s0004t01_lesionsmooth_1.mat does.

[Jane]

Thank you for your time. I have attached this file in this reply.

[Jane]

Actually I have 38912 files in the folder. c0002s0002t01_lesionsmooth_115.mat is one of the files. But i cant upload all due to the limitation of the system.

May I know how can I do to fix this?

Thank you for your time.

Answer 1

[Walter Roberson]

load  c0002s0002t01_lesionsmooth_115.mat
whos

  Name            Size             Bytes  Class          Attributes

  ans             1x59               118  char                     
  cmdout          1x33                66  char                     
  imLesion      192x192            36972  categorical              

The information in the file is already categorical, but your custom read function is trying to use

categorical(labels, [0, 1], {'Background', 'Tumor'})

which would be under the assumption that labels is numeric and that numeric 0 is to be associated with 'Background' and numeric 1 is to be associated with 'Tumor'

uc = unique(imLesion)

uc = categorical
     0 

double(uc)

ans = 1

Your existing imLesion array is categorical with category named 0 -- not category with value 0.

data = categorical(imLesion == '1', [0 1], {'Background', 'Tumor'});
data(1:3,1:3)

ans = 3×3 categorical array
     Background      Background      Background 
     Background      Background      Background 
     Background      Background      Background 

[Jane]

Thank you so much!