1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
|
# docker_render
The `trimesh/trimesh:latest` docker images have a helper script which installs the requirements to use the [LLVMpipe](https://www.phoronix.com/scan.php?page=news_item&px=LLVMpipe-Mesa-19.0-Performance) software rasterizer and XVFB. It's not as fast as a real GPU but it can do ~30FPS with a decent CPU which for some applications is fine.
The advantage of not using a GPU is so you can use it in "normal" cloud instances. Also whether `scene.save_image` is able to render anything ends up being very OS, GPU, and driver specific. With the docker images it should pretty much work on most clients. Software rendering also is way easier than configuring a GPU to be usable in a container (CUDA, EGL, etc).
Take a look at the `Dockerfile`. This will render a PNG of a sphere, probably:
```
# build the image in the current directory
docker build -t renderworker .
# run the example
# `-v`: mount the current directory as a volume
# in the `/output` folder inside the image
docker run -v `pwd`:/output renderworker
```
It should print a bunch of log messages, and something like `rendered bytes: 49803`. You should now see a PNG in the current directory:
```
mikedh@orbital:docker_render$ ls
Dockerfile README.md output.png render.py run.bash
```
It's a sphere!
|
