scene_json_format.txt

We use a simple scene description format for running trained networks on new 
scenes. The scene description uses JSON and looks like this

{
  "walls" : [{"path":"../datasets/models/Box.obj", "translation":[0,0,0]} ],
  "fluids": [{"path":"../datasets/models/Fluid.obj", "velocity":[0.3,0,0.2], "translation":[0,1,0], "start":0, "stop":1, "step":1}]
}

The root object is a dictionary which contains the keys "walls" and "fluids"
which describe the static and the fluid particles respectively. Particles will
be created from obj files with surface sampling (walls) and volume sampling
(fluids). 
Most parameters are self-explanatory. The fluid objects have the
fields "start", "stop" and "step", which define when, how long and at which 
frequency fluid particles will be generated. In the example we generate fluid
particles once in the first simulation frame. To create an emitter that creates
particles every 20 frames until frame 200 one would use 
  "start":0, "stop":200, "step":20