Image to number prediction using CNN regression.

Hello

I have 564 Nifti images, each being 128*128*16*4, as input training data 'imds'. The outputs are 564 continous variables from 0 to 1, which I set as 'imds.Labels'.

I used imagedatastore to combine 'imds' with 'imds.Labels' and then run the regression CNN. But when I run the training, a error suggested that: "the labels of the imagedatastore must be a categorical vector".

Does anyone know how to fix this problem.

Many thanks

Gabriel

[Divya Gaddipati]

Could you share the network you are using ?

[GABRIELZHU ZHU]

Thank you for your response. Actually, I have figured it out, using imds2array().

[hawi teshome]

Hello there, I am facing similar problem. Can you share how you fixed it?

Answer 1

[yanqi liu]

yes，sir，may be use

regressionLayer

as the net output，and use 4-D XTrain, 1-D YTrain to train net