Faster way to make movies than drawing figures

I am trying to create 60 fps movies in matlab, they should only be a few seconds long but drawing each plot takes an excrutiatingly long time. I can't really reduce the fps because things move really quickly and I don't want it to look jerky. Turning the figure visibility off doesn't give any speed up and just stops me knowing how close to being done the process is.

Is there no faster way to make a movie based on plotted data?

F(num_frames) = struct('cdata',[],'colormap',[]);
for frame = 1:num_frames
    cla(ax)
    hold(ax ,'on')
    grid(ax, 'on')
    xlim(ax, [-axlim axlim])
    ylim(ax, [-axlim axlim])
    
    plot(ax, [0, x(frame, 1)], [0, y(frame, 1)])
    
    for i = 1:3
        plot(ax, [x(frame, i), x(frame, i+1)], ...
                 [y(frame, i), y(frame, i+1)])
    end
    drawnow
    F(frame) = getframe(f);
end

Answer 1

[Rik]

You should initialize the graphics objects outside the loop and then modify the properties inside the loop.

The dynamic expansion of F could still be a bottleneck though. Also note that you are using j as an index there, so you can speed this code up by only executing the last frame iteration.

F(num_frames) = struct('cdata',[],'colormap',[]);
cla(ax)
hold(ax ,'on')
grid(ax, 'on')
xlim(ax, [-axlim axlim])
ylim(ax, [-axlim axlim])
h_line1=plot(ax, [0, x(1, 1)], [0, y(1, 1)]);
for i = 1:3
    h_line_i(i)=plot(ax, [x(1, i), x(1, i+1)], [y(1, i), y(1, i+1)]);
end

for frame = 1:240
    set(h_line1,'XData',[0, x(frame, 1)],'YData',[0, y(frame, 1)])
    for i = 1:3
        set(h_line_i(i),...
            'XData',[x(frame, i), x(frame, i+1)], ...
            'YData',[y(frame, i), y(frame, i+1)])
    end
    drawnow
    F(j) = getframe(f);
end

[Conor Marsh]

Hi Rik, thank you for taking the time to answer. I have updated my question's code, I wasn't dynamically expanding F, I missed the j index when making everything readable.

Using your code gives a nice speed up! It was taking 79.202345 seconds to produce a 30 second clip, with your method is takes 48.155147 seconds.

It still feels like this is a long time to wait considering all I'm doing is plotting graphs when I could throw out thousands of frames per second in something like Unity to draw 3D worlds with lighting and shadows etc.

[Rik]

I don't know if getframe is the fastest method to convert your figure to an image. I suspect there are much faster ways to do that, depending on what you want as a result. You are paying for the robustness with time. If I comment out the getframe I get an fps of around 750, meaning that 45 seconds of that 48 are spent in getframe.

figure(1),clf(1)
ax=axes;
hold(ax ,'on')
grid(ax, 'on')
xlim(ax, [-1 1])
ylim(ax, [-1 1])
Nframes=60*30;%30 seconds @ 60 fps
x=rand(Nframes,4);
y=rand(Nframes,4);

h_line1=plot(ax, [0, x(1, 1)], [0, y(1, 1)]);
for i = 1:3
    h_line_i(i)=plot(ax, [x(1, i), x(1, i+1)], [y(1, i), y(1, i+1)]);
end
tic
for frame = 1:(3*240)
    set(h_line1,'XData',[0, x(frame, 1)],'YData',[0, y(frame, 1)])
    for i = 1:3
        set(h_line_i(i),...
            'XData',[x(frame, i), x(frame, i+1)], ...
            'YData',[y(frame, i), y(frame, i+1)])
    end
    drawnow
    %F(j) = getframe(f);
end
t=toc;

clc,fprintf('fps is %.1f (%.1f seconds)\n',Nframes/t,t)

[Conor Marsh]

Exactly what I'm hoping to find! There must be a better way to create a movie...

[Rik]

Untested: some options in export_fig may have a better performance.

Answer 2

[Walter Roberson]

You are doing simple plot() calls, and most of your time is spent capturing the frame.

If you have Computer Vision toolbox, then instead of using getframe, use insertShape https://www.mathworks.com/help/vision/ref/insertshape.html and possibly insertText() as well, to "draw" into an array that you have initialized to your desired background color, after which you can copy the array as your "frame", without doing getframe()

[Conor Marsh]

I was worried the solution wouldn't be simple, it may just be easier for me to export my data and render it using different software. Thank you for the help!

[Walter Roberson]

insertShape() is not a difficult call.

    for i = 1:3
        plot(ax, [x(frame, i), x(frame, i+1)], ...
                 [y(frame, i), y(frame, i+1)])
    end

Are you needing each segment to be a different color? Is that why you are drawing segment by segment instead of in vectorized form?

[Conor Marsh]

It would be ideal if they were different colours yes

[Walter Roberson]

bg = ones(Rows, Columns, 3);      %white

colors = parula(64);              %colors to rotate through
ncolor = size(colors,1);

cidx = 0;
for frame = 1:(3*240)
        I = bg;
        for i = 1 : 3
            cidx = 1 + mod(cidx, ncolor);   %next color from list
            M = [x(frame, i), y(frame,i), x(frame, i+1), y(frame, i+1)];
            I = insertShape(I, 'Line', M, 'Color', colors(cidx,:));
        end
end

except improved to have the other line as well.

[Conor Marsh]

I was hoping to test this today but my admin rights won't let me install the computer vision toolbox... when my IT services get back to me I'll get back to you.

Thank you very much for the help!