Input to Convolution Neural Network

I am using convolution neural network on face images of the size 1024*768*3.

I want to pass these images as input to first input layer of CNN.

what should be size of the 'inputImageLayer' for CNN?

How to decide the size of First layer of CNN?

Answer 1

[bharath pro]

The ImageInputLayer is used to take image inputs when designing a CNN. According to the documentation, it can be called using ImageInputLayer([h w c]), where c= number of chanels. I am assuming for you the height and width of images are 1024 and 768 respectively and c=3. So for this you would need to use ImageInputLayer([1024 768 3]) to instantiate this layer.

[AP]

Thank you bharat pro. I have tried with these dinensions but it is still giving me error ' Not enough input arguments '.

[bharath pro]

Is the input vector a 1*3 row vector? Can you check its dimensions?

[AP]

My image has dimension height 768 width 1024 color representation sRGB.

[bharath pro]

What output do you get when you run size(image)? Also the row vector has to be [768 1024 3] in your case.

[AP]

My output for size(image) is 768 1024 3. Still i am getting error saying 'inputArguments = iParseInputArguments(varargin{:});.

[bharath pro]

Can you share your code here. The problem could be in the other layers as well.

[AP]

Clc;

clear all;

close all;

%read image

a = imread('tryb.jpg');

size(a)

a1 =rgb2gray(a);

%show image

figure, imshow(a1);

imageSize =[768 1024 3];

matlabroot ='D:\MY DESKTOP\IMAGE PROCESSING MATERIAL\database\for nose detect\';

Datasetpath =fullfile(matlabroot ,'cnn2','Dataset1');

Data = imageDatastore(Datasetpath,'IncludeSubfolders',true,'LabelSource','foldernames');

layers = [

imageInputLayer (imageSize)

covolution2dLayer([5 20])

reluLayer

maxPooling2dLayer(2,'stride',2)

covolution2dlayer([5 20])

reluLayer

maxPooling2dLayer(2,'stride',2)

fullyConnectedLayer(2)

softmaxLayer

ClassiifcationLayer()];

options = trainingOptions('sgdm','maxEpoch',15,'initialLearningrate , 0.001');

convnet = trainNetwork (Data,layers,options);

output = classify (convnet ,a);

[bharath pro]

change your layers to this:

layers = [
         imageInputLayer(imageSize,'Normalization','none')
         convolution2dLayer(5,20)
         reluLayer()
         maxPooling2dLayer(2,'Stride',2)
         convolution2dLayer(5,20)
         reluLayer()
         maxPooling2dLayer(2,'Stride',2)
         fullyConnectedLayer(1)
         softmaxLayer
         classificationLayer()];
  

Also your options seem to be wrong. Please change that.