Deduplicate Geometry#

Key: deduplicateGeometry

This replaces multiple duplicate geometric prims in a scene with a single prim plus references/instances to it. Since a referenced prim uses less memory than a full duplicated prim, this can reduce both system memory and VRAM. It is only effective when there are prims that are identical but not already instanced, so it may have no effect on a scene that is already well instanced.

A fuzzy comparison mode is also available: its similarity measure is independent of tessellation and based on relative shape deviation, with CPU and GPU implementations. The operation deduplicates point-based geometry (meshes, basis curves, etc.); in fuzzy mode only meshes are supported.

This is mesh-level deduplication: it matches individual gprims, not whole sub-trees. To collapse duplicate assemblies (entire prim hierarchies), run Deduplicate Hierarchies first, then this operation to catch any remaining loose duplicate meshes.

Matching controls#

tolerance (default 0.001, stage units) is the position tolerance for considering two meshes equal; use 0 to require exact matches. fuzzy enables shape-based matching; allowScaling lets uniformly scaled copies match; considerDeepTransforms (default true) accounts for the full world transform when comparing. minimumDuplicates (default 2) sets how many copies must exist before a prototype is created. ignoreAttributes excludes named attributes from the comparison.

Starting configurations#

Exact instancing (default method):

[{"operation": "deduplicateGeometry", "duplicateMethod": 2, "tolerance": 0.001}]

Fuzzy (tessellation-independent) matching:

[{"operation": "deduplicateGeometry", "duplicateMethod": 2, "fuzzy": true, "allowScaling": true}]

Arguments#

Geometry to De-duplicate#

Optional list of prim paths to consider

  • Name: meshPrimPaths

  • Type: [string]

  • Default Value: []

Tolerance#

Acceptable point position change during deduplication. The value is a stage unit in worldspace

  • Name: tolerance

  • Type: float

  • Default Value: 0.001

Method#

Method used to conform meshes that are duplicates

  • Name: duplicateMethod

  • Type: int

  • Default Value: 2

  • Enum Values:

    • 0: Copy Values

    • 1: Reference

    • 2: Instanceable Reference

    • 3: Set Attribute

    • 4: Point Instancer

Point Instancer Location#

Where to author the PointInstancer for each duplicate set

  • Name: pointInstancerLocation

  • Type: int

  • Default Value: 0

  • Enum Values:

    • 0: Common Root

    • 1: Custom Path

Parent Path#

Prim path to author the PointInstancer under. Created as an Xform if it does not exist.

  • Name: pointInstancerParentPath

  • Type: str

  • Default Value: ""

Minimum Duplicates#

Minimum number of duplicates a set must contain for a PointInstancer to be created. Sets with fewer duplicates are left untouched.

  • Name: minimumDuplicates

  • Type: int

  • Default Value: 2

  • Min Value: 2.0

Ignore Attributes#

Optional list of attributes to ignore. This list can be explicit attributes, or if ending with a ‘:’ can ignore namespaces.

  • Name: ignoreAttributes

  • Type: [string]

  • Default Value: []

Fuzzy mode#

When enabled, uses shape comparison to find duplicates that differ in tessellation or have baked-in point offsets

  • Name: fuzzy

  • Type: bool

  • Default Value: False

Allow Scaling#

When enabled, fuzzy comparison will factor out uniform scaling

  • Name: allowScaling

  • Type: bool

  • Default Value: False

Consider Deep Transforms#

Look for duplicates where the points values have been uniformly transformed

  • Name: considerDeepTransforms

  • Type: bool

  • Default Value: True