Node groups based on the new Singular Value Decomposition (SVD) node added in Blender 5.1. Requires Blender 5.1 or above!

Oriented Bounding Box

This node computes a bounding box rotated to fit the geometry better. This method is also known as Principal Component Analysis (PCA).

It's useful for generating proxy geometry, or to move geometry back to origin so it can be adjusted in rest space.

For more information, see the demo video, this video by CGMatter and this article by Andy Nicholas.

Estimate Transform

This node takes two sets of geometry, the Source Geometry and Target Geometry, then computes the best possible transform to move one to the other.

It assumes both set of points have the same count and are in the same order. If this is not the case, you should look into Iterative Closest Point (ICP).

It's equivalent to the "Extract Transform" node in Houdini. The Kabsch-Umeyama algorithm is used for translation, rotation and uniform scale.

For more information, see the demo video, this article by Nghia Ho and my article about Extract Transform.

Outer Product 3x3

Used by Oriented Bounding Box and Estimate Transform. Computes a 3x3 outer product of two 3D input vectors, storing the result in the 3x3 part of the output matrix.

Outer Product 4x4

Used by Estimate Transform. Computes a 4x4 outer product of two 3D input vectors, assuming the W component of both is 1.

Accumulate Matrix 3x3

Used by Oriented Bounding Box and Estimate Transform. Sums all the values of a 3x3 matrix field using Attribute Statistic, used to compute a covariance matrix.

Accumulate Matrix 4x4

Used by Estimate Transform. Sums all the values of a 4x4 matrix field using Attribute Statistic, used to compute a covariance matrix.