Frame per second computingfor GPU code generation

Hello all

In Matlab documentation "Code Generation for Deep Learning Networks", I can not understand why they used reccurssive technique to compute frame per second, this is the code

% Connect to a camera
camera = webcam; 
% The labels with top 5 prediction scores are 
% mapped to corresponding labels
net = mobilenetv2();
classnames = net.Layers(end).ClassNames;
imfull = zeros(224,400,3, 'uint8');
fps = 0;
ax = axes;
while true
    % Take a picture
    ipicture = camera.snapshot;       
    % Resize and cast the picture to single        
    picture = imresize(ipicture,[224,224]);  
    % Call MEX function for MobileNet-v2 prediction
    tic;    
    pout = mobilenetv2_predict(single(picture));
    newt = toc;  
    % fps 
    fps = .9*fps + .1*(1/newt);
    % top 5 scores
    [top5labels, scores] = getTopFive(pout,classnames);  
    % display
    if isvalid(ax)
        dispResults(ax, imfull, picture, top5labels, scores, fps);   
    else
        break;
    end
end
end

Why they have not just used instead of the reccurssive equation

fps =1/newt;

Thanks in advance for your answers

Answer 1

[Joss Knight]

This is a smoothing filter to eliminate noise, to ensure it gives a good result from the first frame you could set it to 1/newt on the first iteration.