Rasterizer#

The default renderer: fast, GPU-accelerated rasterization for real-time viewing, control loops, and reinforcement-learning rollouts. It is also the backend the interactive viewer always uses. Name it explicitly with gs.Scene(renderer=gs.renderers.Rasterizer()), though it is already the default when renderer is omitted.

The options class takes no parameters. Rasterizer behavior such as shadows, lights, and per-environment isolation (env_separate_rigid) is configured on gs.options.VisOptions, not on the renderer. See Rendering for adding cameras and reading back images.

Options#

class genesis.options.renderers.Rasterizer None[source]#

Rasterizer renderer. This has no parameter to be configured.

Note

You can set which renderer to use for cameras, but the interactive viewer always uses the rasterizer rendering backend. If you want to configure properties like shadow, lights, etc., you should use gs.options.VisOptions instead.

Implementation#

class genesis.vis.rasterizer.Rasterizer(viewer, context)[source]#

Bases: RBC

build()[source]#
add_camera(camera)[source]#
update_camera(camera)[source]#
remove_camera(camera)[source]#
render_camera(camera, rgb=True, depth=False, segmentation=False, normal=False, *, split_envs)[source]#

Render a camera. With ‘split_envs’, the environments are rendered one by one from the pose the camera holds in each, and stacked; otherwise one image is rendered of the environments laid out side by side.

update_scene(force_render: bool = False)[source]#
destroy()[source]#
property viewer#
property offscreen#

See also#