Deep Learning Layers, incorrect output

I have created the following layer. Outside the layer, I have tried that if I have input1 size of [68,68,1], input2 size of [68,1,3] and input3 size of [68,17,3], I should be able to get back Z1 of size [17,17,1] and Z2 of size [17,1,3]. But instead when I try to analyze the layer, it gave me back scalars for Z1 and Z2 (out1 and out2 both equals to 17), I don't know what I have done wrong.

classdef SpectralPoolingLayer < nnet.layer.Layer

% Example custom weighted addition layer.

properties

end

properties (Learnable)

% Layer learnable parameters

% Scaling coefficient

end

methods

function layer = SpectralPoolingLayer(numInputs,numOutputs,name)

% layer = weightedAdditionLayer(numInputs,name) creates a

% weighted addition layer and specifies the number of inputs

% and the layer name.

% Set number of inputs.

layer.NumInputs = numInputs;

layer.NumOutputs = numOutputs;

% Set layer name.

layer.Name = name;

% Set layer description.

layer.Description = "SpectralPooling of " + numInputs + ...

" inputs";

% Initialize layer weights.

end

function [Z1,Z2] = predict(~, X1,X2,X3)

% Z = predict(layer, X1, ..., Xn) forwards the input data X1,

% ..., Xn through the layer and outputs the result Z.

% Initialize output

A = X1(:,:,1);

P = X3(:,:,1);

sz3 = size(X3);

Z1 = zeros([sz3(2),sz3(2),1],'like',X3);

sz = size(X2);

Z2 = zeros([sz3(2),1,sz(3)],'like',X2);

%start Implementing

Z1(:,:,1) = P'*A*P;

Z1 = max(0,Z1);

for i = 1:sz(3)

Z2(:,1,i) = P'*X2(:,1,i);

end

end

end

end