How to change the image input layer in GoogleNet?

Hello everyone, I'm working on designing of an age estimation system and I started working using GoogleNet architecture for CNN as transfer learning. my question is How can I change the input image layer from [224 224 3] (RGB) to [250 250 1] for gray scale image? I think even if I can change the input image size, the configuration of the network as pretrained will not keep configured! what can i do? when I try to change the image input layer, I get error as below >>net = googlenet; >> lgraph=layerGraph(net); >>lgraph.Layers(1)=((imageInputLayer([250 250 1],'Name','input'))); You cannot set the read-only property 'Layers' of LayerGraph.

[Kevin Chng]

How about resize your input Image instead of change the layer properties?

[Rami Sarhan]

I understand, but i’m using gray scale image

[Kevin Chng]

As what I know, you can't train the input size as googLenet train the model based on RGB. You may need to train your own deep learning model from scratch, or transfer learning using AlexNet.

Or consider to convert your grayscale to RGB

[Mrutyunjaya Hiremath]

Load Pretrained Network in API - Deep Network Designer. delete and add new layers and analyze it

Answer 1

[Mrutyunjaya Hiremath]

Using API - Deep Network Designer

[SHURGIL UR]

net = alexnet; change this into googlenet