Sample count 3D positions from clusters gaussian blobs inside a
[-bounds, bounds]^3 volume — the shared spatial scaffold for the
points and gsplats generators (real microscopy / astronomy point
sets are clumpy, and clustered depth structure is what makes
depth-sorted blending measurably order-dependent).
Cluster centers are uniform in the volume; per-cluster σ is
2–8 % of bounds. Deterministic for a given (rand-stream, count, clusters, bounds).
Also returns the empirical min/max corner of the sampled positions
(gaussians have unbounded tails, so consumers needing a bounding box
must use the measured one, not ±bounds).
Sample
count3D positions fromclustersgaussian blobs inside a[-bounds, bounds]^3volume — the shared spatial scaffold for the points and gsplats generators (real microscopy / astronomy point sets are clumpy, and clustered depth structure is what makes depth-sorted blending measurably order-dependent).Cluster centers are uniform in the volume; per-cluster σ is 2–8 % of
bounds. Deterministic for a given(rand-stream, count, clusters, bounds).Also returns the empirical min/max corner of the sampled positions (gaussians have unbounded tails, so consumers needing a bounding box must use the measured one, not
±bounds).