The input images for activations must be of size equal to or greater than [224 224 3].

my code is:

augmentedTrainset=augmentedImageDatastore(imagesize,...
    trainset,'ColorPreprocessing','rgb2gray');
augmentedTestset=augmentedImageDatastore(imagesize,...
    testset,'ColorPreprocessing','rgb2gray');
featureLayer = 'fc1000';
trainingFeatures = activations(net, augmentedTrainset, featureLayer, ...
    'MiniBatchSize', 39, 'OutputAs', 'columns');

[224 224 3] image size

And I take an error as "The input images for activations must be of size equal to or greater than [224 224 3]."

Why I'm taking this error, how can I fix it?

[Geoff Hayes]

aitali - does your test set contain any images that are smaller than 224x224x3?

[aitali hassna]

yes the size of all images i test is 224x224x3

Answer 1

[Sai Bhargav Avula]

Hi,

Is your image [224 224 3] or of different order?

And try the below code once

augmentedTrainset=augmentedImageDatastore(imagesize,...
    trainset);
augmentedTestset=augmentedImageDatastore(imagesize,...
    testset);
featureLayer = 'fc1000';
trainingFeatures = activations(net, augmentedTrainset, featureLayer, ...
    'MiniBatchSize', 39, 'OutputAs', 'columns');

[aitali hassna]

Yes the size of my image is [224 224 3]

thanks for your answer , doesn't work for me

[Sai Bhargav Avula]

can you try by removing the 'ColorPreprocessing' name value argument.

The ColorPreprocessing in this case is reducing the channnels from 3 to 1. If all all the images have the same number of channels(i.e 3 in your case) then no processing is required. It is generally used when you have a mix of gray scale and color images(specifically RGB) and to make sure that all output images have the same number of channels . I have updated the above answer accordingly.

Hope that helps!

[aitali hassna]

Thanks sir!

that works for me

augmentedTrainset=augmentedImageDatastore(imagesize,...

trainset,'ColorPreprocessing','gray2rgb');

augmentedTestset=augmentedImageDatastore(imagesize,...

testset,'ColorPreprocessing','gray2rgb');

featureLayer = 'fc1000';

trainingFeatures = activations(net, augmentedTrainset, featureLayer, ...

'MiniBatchSize', 39, 'OutputAs', 'columns');

[aitali hassna]

augmentedTrainset=augmentedImageDatastore(imagesize,...
    trainset,'ColorPreprocessing','gray2rgb');
augmentedTestset=augmentedImageDatastore(imagesize,...
    testset,'ColorPreprocessing','gray2rgb');
featureLayer = 'fc1000';
trainingFeatures = activations(net, augmentedTrainset, featureLayer, ...
    'MiniBatchSize', 39, 'OutputAs', 'columns');

[Sai Bhargav Avula]

The ColorPreprocessing name value can be removed. As all the images already has 3 channels no processing occurs.

[aitali hassna]

yes exactly sir that work for me soo but i didn't py attention that i have some image has 1 channel so the fault that i make 'rgb2gray' in place of 'gray2rgb'

thanks sir

for your help