Optimize Cutout Geometry Nodes

- Refactor Depth Cutout to reuse the projected outline field, reducing redundant boundary calculations and improving performance.
- Derive the center field from shared low-frequency fields, eliminating an unnecessary blur operation.
- Implement triangle-only strip cleanup for Depth Cutout, as the input is confirmed to be triangulated.
- Short-circuit zero-thickness branches in both Cutout and Depth Cutout to avoid unnecessary geometry generation.
- Standardize internal attribute naming with a new `_o_` prefix and implement wildcard cleanup for temporary attributes.
- Update relevant tests to ensure functionality remains consistent and validate changes against existing behavior.
This commit is contained in:
2026-09-16 08:42:13 +08:00
parent cf512dc013
commit 7f878a2f3f
22 changed files with 584 additions and 135 deletions

View File

@@ -130,7 +130,7 @@ def build_image_depth_plane_group(image_plane_group):
surface = project_depth_surface(
group, geometry, camera,
scale_input.outputs["Uniform Scale"], depth_input.outputs["Reference Depth"],
depth_input.outputs["Depth Scale"], split_input.outputs["Depth Split"],
depth_input.outputs["Depth Scale"],
)
surface = smooth_cut_boundary(
group, surface, cut_vertex, protected=original_boundary, outline_scale=1.0,