search

facebookresearch / sound-spaces

A first-of-its-kind acoustic simulation platform for audio-visual embodied AI research. It supports training and evaluating multiple tasks and applications.
https://soundspaces.org
Creative Commons Attribution 4.0 International
322 stars 55 forks source link

channel type #95

Open yzjqh7 opened 1 year ago

yzjqh7 commented 1 year ago

Hi,

Great work on the simulation platform! I was running the simulation on a 3D mesh I created, but I figured that when I use Mono channel for simulation, the output does not contain anything. See message below. When I use binaural channel, all things worked fine. Is this a bug, or I missed something?

My other question is does the agent itself has some structure that might modify the signal? Or the agent is transparent for simulation.

image
ChanganVR commented 1 year ago

Hi @yzjqh7, there is a bug in the binaries and it's being fixed right now. For now, to render mono, set the microphone type to Ambisonics and set the number of channels to 1 and the returned audio is exactly mono.

The agent itself is seen as transparent for audio propagation unless the microphone type is binaural, in which case the human head shape is taken into account.