My image size is of [566 804 3], what are the useful convolution filter sizes? How can I predict them? Every where I just given the same filter size and same number of filters?

1 view (last 30 days)

GUDIVADA ROKESH KUMAR on 16 Feb 2019

0
Link

Direct link to this question

https://au.mathworks.com/matlabcentral/answers/445315-my-image-size-is-of-566-804-3-what-are-the-useful-convolution-filter-sizes-how-can-i-predict-the

layers=[...
    imageInputLayer([566 804 3])
    convolution2dLayer(50, 20)
    reluLayer
    crossChannelNormalizationLayer(2)
    maxPooling2dLayer(5,'stride',2,'padding',2)
    convolution2dLayer(50, 20)
    reluLayer
    crossChannelNormalizationLayer(2)
    maxPooling2dLayer(5,'stride',2,'padding',2)
    convolution2dLayer(50, 20)
    reluLayer
    crossChannelNormalizationLayer(2)
    maxPooling2dLayer(5,'stride',2,'padding',2)
    convolution2dLayer(50, 20)
    reluLayer
    convolution2dLayer(50, 20)
    reluLayer
    convolution2dLayer(50, 20)
    reluLayer
    maxPooling2dLayer(5,'stride',2,'padding',2)
    fullyConnectedLayer(2)
    softmaxLayer
    classificationLayer()]