Skip to content

PyVista Dataset Ideas for Holograms

Source: pyvista.examples.downloads (checked directly against the installed package, PyVista 0.48.4 -- .venv/lib/python3.12/site-packages/pyvista/examples/downloads.py)

A survey of PyVista-reachable datasets worth feeding into the LFD / HLD pipeline, focused on subjects with real depth structure -- the thing that actually makes a quilt fuse into something worth looking at.

Geography / topography

These have strong, legible elevation relief -- good raw material for parallax. Check the disparity budget once you pick a camera setup; global-scale datasets can blow it if the near/far range isn't tamed.

Function Type What it is Notes
examples.download_crater_topo() + download_crater_imagery() ImageData + Texture Mt. Ruapehu (New Zealand) crater DEM with a draped aerial GeoTIFF PyVista's own "Topographic Map" tutorial (Ruapehu_mag_dem_15m_NZTM.vtk -- confirmed from the source file name, NZTM = NZ Transverse Mercator). A real crater bowl plus a photo-realistic texture -- good first candidate, controlled depth range.
examples.download_st_helens() ImageData Mt. St. Helens post-eruption DEM dataset.plot(cmap="gist_earth"). Terrain relief alone, no texture -- simpler than the crater pair above.
examples.download_topo_global() PolyData Whole-Earth topography + bathymetry, as a sphere Full globe, pole-to-trench depth range. Striking, but the depth range is huge -- expect to need a narrow view cone or a tight focal-plane placement.
examples.download_topo_land() PolyData Land-only global elevation clim=[-2000, 3000], cmap="gist_earth". Same globe without the ocean floor -- cleaner, smaller depth budget than the full version.
examples.download_damavand_volcano() ImageData Mt. Damavand (Iran) volumetric data Isosurface or volume-render. A single conical peak is easy to reason about depth-wise -- good for a first hologram test.

Zero-download option for pipeline iteration: examples.load_random_hills() is synthetic rolling terrain, useful for testing the quilt/sweep code before pointing it at a real multi-MB DEM.

Brain volumes

Human -- native to PyVista

Checked directly against the installed package: no dataset named "mouse" or "mouse brain" exists in pyvista.examples. What PyVista ships is human, and both are volume-render-ready ImageData with no download plumbing needed beyond the examples.download_*() call itself:

Function What it is Notes
examples.download_brain() Classic VTK brain.vtk volume -- a human head MRI dataset.plot(volume=True). Used in PyVista's own volume-rendering, slicing, depth-peeling, and moving-isovalue tutorials -- well-trodden, predictable behavior.
examples.download_brain_atlas_with_sides() avg152T1_RL_nifti.nii.gz -- the MNI152 averaged human brain template, left/right labeled dataset.slice(normal="z").plot(cpos="xy"). An averaged brain (152 subjects) rather than one individual's scan -- smoother, less idiosyncratic anatomy than download_brain().

Both are good, zero-friction volume-render subjects for testing the LFD volume-sweep path before moving to the much larger Allen mouse data below.

Mouse -- the real volume: Allen Institute CCFv3

The Allen Mouse Brain Common Coordinate Framework is a real 3-D mouse atlas built from 1,675 C57BL/6J mice, distributed as plain NRRD files -- no API key, no AllenSDK dependency required:

http://download.alleninstitute.org/informatics-archive/current-release/mouse_ccf/average_template/average_template_50.nrrd
  • Resolutions: 10 / 25 / 50 / 100 µm isotropic (average_template_{res}.nrrd). Start with 50 µm -- the finer volumes get large fast.
  • There's also a labeled version at the same resolutions: mouse_ccf/annotation/ccf_2017/annotation_50.nrrd -- same shape, but every voxel is a brain-region ID instead of grayscale intensity. This is probably the more striking hologram candidate: a segmented, colorable volume rather than plain grayscale.
  • PyVista reads .nrrd natively (pv.read() / pv.NRRDReader), so it drops into an ImageData volume exactly like download_brain() does -- no extra plumbing needed in lfd.py / hld.py.
import pyvista as pv

vol = pv.read("average_template_50.nrrd")   # -> pyvista.ImageData
vol.plot(volume=True, cmap="bone")

scripts/render_pyvista_hologram.py mouse-brain downloads and caches this for you, in the platform's native per-user cache directory (~/Library/Caches/quiltwright/allen_ccf on macOS, $XDG_CACHE_HOME/quiltwright/allen_ccf on Linux) -- see tvb-data.md for the full resolution order, which is shared with the other runtime downloads. These volumes run from ~60 MB at 100 µm to well over a gigabyte at 10 µm, so set $QUILTWRIGHT_ALLEN_CACHE if you would rather keep them on another disk.

Allen Institute data is free for non-commercial use under their terms -- see the data license before any redistribution.

Other strong-depth candidates worth a look

Not geography, but structurally similar in that they have real self-occlusion and depth layering rather than a flat relief:

  • examples.download_frog() / examples.load_frog_tissues() -- classic segmented full-body CT scan (frog), colorful multi-organ volume.
  • examples.download_whole_body_ct_male() / _female() -- human whole-body CT, much larger and more detailed than the frog.
  • 99-advanced/gyroid, 99-advanced/atomic_orbitals, 99-advanced/sphere_eversion -- abstract math surfaces with deep self-occlusion; useful as parallax stress tests outside the "real world scan" category.

Sources