I now want to get camera images with an API “get_camera_image” without launching viewer. (As I’m using isaac gym via ssh, so when I tried to launch viewer, I ended up with getting the following error. I also used “get_camera_image” without launching viewer, and I naturally got all-0 image.)
It doesn’t look like possible to get the correct image without viewer, but if you know the way to do that, glad if you could share.
error message
Exception: Failed to create viewer
Not too familiar with isaacgym myself but the graphics.py, multiple_camera_envs.py and interop_torch.py examples are able to produce images without a viewer so may be worth checking out.
Thanks a lot for your replay.
I’m going to check those code and see what’s going on there.
Thank you !!
I now found that if I use write_camera_imgI can see images without a viewer, but I cannot do the same thing with get_camera_img. More precisely, I can get an image, but I get completely black image.
Does anybody have any ideas on what’s causing this problem?
You can take a look at the graphics.py example for using get_camera_image and Pillow to export the depth images. It will also work in headless mode without the viewer when running with --headless.
The graphics.py example uses a CPU pipeline. So will get_camera_image work in a headless mode with a GPU pipeline?
Will toggling the collision filter could show/hide robots in front of a particular actor?