I was using CNN3D model. But its total uncelar how the data should be loaded for network to train.

I was using CNN3D model. But its total uncelar how the data should be loaded for network to train. It can either be table, or array. My image size is 22x198x153x3 and i have 100 sample. another way of looking at it is i have 22 3d matrices stacked over other.

I converted it to struct2table of 100 rows, not working, it says too many arguments.

Please advise how to load data for training in cnn3d.

[awezmm]

Can you give the exact command you are running and the errors associated with it?

[Sujit Roy]

ok.

I have 100 mat data file named like tube001.mat which has dimension of 22x198x153x3, where 22 is the depth 196 is the length and 153 is width and 3 is for rgb.

So I loaded them all in traindata which is now of size 1x100 struct; 100 representing the number of file. then i converted them to table as network requirement by struct2table(traindata).

similarly I have labels if 1x100 double which is converted to categorical.

for cnn training

image3dInputLayer(size(traindata.tube{1,1}))

and then when i run

net = trainNetwork(traindata, categorical(labels),layers,options);

error @@@@

Error using trainNetwork (line 165)

Too many input arguments.

Caused by:

Error using trainNetwork>iParseInputArguments (line 290)

Too many input arguments.

@@@@

Answer 1

[Shivam Sardana]

You can use imageDatastore to store image data.

[Sujit Roy]

You cannnot that is for 2d not 3d

[Binu]

you can use imageDataStore for 3d but you'll have to save your volumes as .mat files