Screenshot with docker

[hhoangg]

I deploy source with docker, server Linux, have not GPU
Like the title, I want to use a screenshot function to render the mesh texture layer, but not working
is there any way to export images without GPU?
this my source

from vedo import Mesh, settings, show

settings.screenshotTransparentBackground = True
mesh = Mesh("test.obj").texture("test-design.png").forceOpaque()

x0,x1 = mesh.xbounds()
y0,y1 = mesh.ybounds()
size = ((x1-x0)*5, (y1-y0)*5)

plt = show(mesh, size=size, zoom='tight', interactive=False)

plt.screenshot("result.png")

Linked issue

[antmatyjajo]

Hi,

It's quite possible, I am using vedo in this way.

You need to set everything up for offscreen rendering: there are two main ingredients

1. Vedo should be set to render in offscreen mode
2. Guest OS in the docker container needs the relevant libraries installed (e.g., see here and here. Specifically in this example we need the Mesa openGL and GLX extensions, and Xvfb to act as a virtual screen.

(I think it's maybe also possible to use OSMesa offscreen driver directly, but that requires a custom build of VTK as far as I know...)

Here is what I am using (NB I am not a docker expert so this is not necessarily optimal or anything...)

1. Dockerfile

FROM python:3.8-slim-bullseye

RUN apt-get update -y \
  && apt-get install libgl1-mesa-dev libgl1-mesa-glx xvfb -y --no-install-recommends \
  && apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false \
  && rm -rf /var/lib/apt/lists/*

RUN pip install vedo \
  && rm -rf $(pip cache dir)

RUN mkdir -p /app/data

WORKDIR /app/
COPY test.py set_xvfb.sh /app/
ENTRYPOINT ["/app/set_xvfb.sh"]

2. set_xvfb.sh

#!/bin/bash
set -x
export DISPLAY=:99.0
Xvfb :99 -screen 0 1024x768x24 > /dev/null 2>&1 &
#sleep 3
set +x
exec "$@"

3. test.py

from vedo import Sphere, Plotter, settings

settings.screenshotTransparentBackground = True

vp = Plotter(interactive=False, offscreen=True)

s = Sphere(pos=[-5, 0, 0], c="r")
vp.show(s)

vp.screenshot("./data/out.png", scale=2)

Then you can

1. $ docker build -t vedo-test-local .
2. $ docker run --rm -v /some/path/output:/app/data vedo-test-local python test.py (directory /some/path/output needs to exist)
3. There should be an "out.png" in the output directory.

Probably a good idea to set specific versions for your packages etc. in the dockerfile to avoid surprises later if vedo API changes or something.

[hhoangg]

@antmatyjajo thanks for the help, but server RAM is not auto-clear when the screenshot function done

[antmatyjajo]

Sure, no problem.
About the RAM, I don't understand what you mean. docker run --rm ... should both stop and then delete the container once the function has finished running, is that not happening?

[hhoangg]

Sure, no problem. About the RAM, I don't understand what you mean. docker run --rm ... should both stop and then delete the container once the function has finished running, is that not happening?

thank you for your help, I did it, thank you very much, if there is any other problem I will create a new issue :D