How to change input size of neural network from 3d to 2d?

8 views (last 30 days)
Jahandad Baloch
Jahandad Baloch on 30 May 2018
Commented: Jahandad Baloch on 31 May 2018
I have to classify 2d images using googlenet model of Neural Network Toolbox, here the input layer size is 224x224x3 but I need it to classify 224x224x1. How to change the size of input layer and what else would be needed to change after this change in input size?

Answers (1)

Majid Farzaneh
Majid Farzaneh on 30 May 2018
Edited: Walter Roberson on 30 May 2018
Hi,
You can use the gray-scale image instead of RGB image. Use this command to make gray-scale from RGB:
Ig=rgb2gray(I)
Where I is your input image.
Also you can use one of layer of image:
R=I(:,:,1);
or
G=I(:,:,2);
or
B=I(:,:,3);
  1 Comment
Jahandad Baloch
Jahandad Baloch on 31 May 2018
Thank you very much Majid and Walter. The problem is, I already have gray-scale 2D images but googlenet model by Neural Network Toolbox accepts data in 3D. Should I introduce third dimension by converting gray-scale to RGB? Wouldn't this extra dimension reduce the performance of algorithm?

Sign in to comment.

Products


Release

R2017b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by