On 9/5/19 6:16 AM, Xavier Cardil wrote:
> Thank you for your responses.
> It took me a day to realize that this might be actually the best
> solution, as plotting to memory will be way faster than writing to disk.
> As long as we can plot to memory via plsmema( ) and then retrieve it,
> it should be possible to convert the plot to a Numpy array ?
> Keras takes images as arrays as per our requirement. In matplotlib this
> is done behind the covers with Pillow + Numpy if I'm not wrong.
> Pillow can store images as arrays in memory, so it's similar to what
> plsmema( ) does. I mentioned RGBA encoded string instead of Numpy array
> because I believe Pillow stores images in memory as RGBA strings, and I
> was trying to find a replacement for the whole procedure ( matplotlib is
> terribly slow )
>
> It would be great to hear more comments from you about this, thanks !
>
Here is an example (using the "memqt" driver, but "memcairo" should also
work).
#!/usr/bin/env python3
from PIL import Image
import numpy
import plplot
width = 480
height = 320
plot_buffer = numpy.zeros((height, width, 4), numpy.uint8)
plplot.plsmema(width, height, plot_buffer)
plplot.plstart("memqt", 1, 1)
plplot.plenv(0, 360, 0, 90, 0, 2)
plplot.plend1()
plot_image = Image.frombytes("RGBA", (width, height), plot_buffer)
plot_image.save("image.png")
The final plot is available in the plot_buffer numpy array object, or in
the plot_image PIL/Pillow Image object.
-Hazen
_______________________________________________
Plplot-general mailing list
Plplot-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/plplot-general
