cnnTestData.m

function [acc, e] = cnnTestData(cnn, VX, VY, numImages)
    % Validate CNN Accuracy
    %   VData: validation data, [x-dim, y-dim, channel-num, data-count]
    %   VLabel: validation label, [1, data-count]
    %   numImages: number of images that want to validate

    cnn.to.test = 1;

    if cnn.to.useGPU == 1
        images = gpuArray(single(VX(:, :, :, 1:numImages)));
        mb_labels = gpuArray(single(VY(:, 1:numImages)));
        cnn = cnnFeedForward_GPU(cnn, images);
    else
        images = VX(:, :, :, 1:numImages);
        mb_labels = VY(:, 1:numImages);
        cnn = cnnFeedForward(cnn, images);
    end

    % if to.PCAflag==1
    %             for iLayer=1:cnn.LNum
    %                 if cnn.Layers{iLayer}.type==9
    %                     fltLayer=iLayer;
    %                     break;
    %                 end
    %             end
    %             OptData=cnnFilter(images, cnn.Layers{fltLayer});
    %         else
    %             OptData=[];
    %         end

    [~, preds] = max(cnn.OutData{cnn.LNum}, [], 1);
    e = (preds == mb_labels);
    acc = gather(sum(preds == mb_labels) / numImages);

end