All pull requests

Pull requests

• spiral: orient winding-inference rays before using their crossing levels

  #2001·rambouhh opened Oct 8, 2026·1 task done·

  ·
  1

• lasagna: benchmark flatten_fast vs flatten_long on real tifxyz; float32 forward-flatten export

  #2000·blackemberlabs opened Oct 7, 2026·1 task·

  ·
  1

• optimized_inference: keep each S3 surface volume's disk cache separate

  #1999·tencur opened Oct 7, 2026·1 task done·

  ·
  2

• optimized_inference: prepare keeps the signal of 16-bit layer TIFFs

  #1998·tencur opened Oct 7, 2026·1 task·

  ·
  4

• optimized_inference: do not re-apply the layer window to the zarr the prepare step cropped

  #1997·tencur opened Oct 7, 2026·1 task done·

  ·
  10

• grow_track_graph: optional support-based final mask — half the flattening CPU for ~98% of the coverage

  #1996·ibarapascal opened Oct 7, 2026·

  ·
  1

• vc_render_tifxyz: --gpu samples the volume on a CUDA device, byte for byte the CPU render

  #1995·ubc-nvining opened Oct 7, 2026Contributor·

  ·
  1

• volume-registration: check transform matrices against their own landmarks (#1843)

  #1994·neg-0 opened Oct 7, 2026·1 task done·

  ·
  1

• surface_preflight: hint whether a surface needs --flip-normals (refs #1648)

  #1993·alexxxmur opened Oct 6, 2026·

  ·
  1

• Windows: embed a UTF-8 application manifest in every executable (fixes #1979)

  #1992·alexxxmur opened Oct 6, 2026·

  ·
  1

  1

• lasagna: blocking copy of the cyl_outside field to non-CUDA devices

  #1990·AndreasHad04 opened Oct 6, 2026·1 task done·

  ·
  1

• vesuvius.predict: --overlap is the overlap, not the stride

  #1989·AndreasHad04 opened Oct 6, 2026·1 task done·

  ·
  1

• vesuvius.predict: run on Apple Silicon (MPS) instead of failing at model load

  #1988·AndreasHad04 opened Oct 6, 2026·1 task done·

  ·
  1

• merge-ink-pipelines: reject multi-page TIFF labels

  #1987·woahwhattheheck opened Oct 6, 2026Contributor·

  ·
  1

• Fix Windows ThreadPool shutdown hang

  #1986·woahwhattheheck opened Oct 6, 2026Contributor·

  ·
  1

  2

• vc_flatten: reject tifxyz scale mismatches before raster allocation

  #1985·woahwhattheheck opened Oct 6, 2026Contributor·

  ·
  1

  1

• self_render_tifxyz: fuse depth projection into one Numba pass (~30–40% less CPU, identical output)

  #1984·ToukoUrsin opened Oct 6, 2026·1 task done·

  ·
  1

• docs: point Linux users to the release AppImage; fix the VC3D tutorial's Docker tag

  #1982·Bullo27 opened Oct 5, 2026Contributor·1 task done·

  ·
  1

• find_valid_patches: plain zarr arrays fall back to full resolution (fixes #1970)

  #1980·DeMarlon opened Oct 5, 2026·1 task done·

  ·
  1

• vc_obj2tifxyz_legacy: size the grid by the median UV-to-3D scale, not the mean

  #1978·evilaliv3 opened Oct 5, 2026Contributor·1 task done·

  ·
  1

  2

• volume-cartographer: use @loader_path rpaths for Python bindings on macOS

  #1975·lucasjbyoung opened Oct 5, 2026·1 task done·

  ·
  1

• InpaintSurface: seed holes with the harmonic fill and take the grid step as the spacing target

  #1972·evilaliv3 opened Oct 4, 2026Contributor·1 task done·

  ·
  1

• Native 3D ink training: work only on the windings a crop meets (identical samples, prepared 6-13x faster than with #1952 on PHerc. Paris 4)

  #1971·Bullo27 opened Oct 4, 2026Contributor·1 task done·

  ·
  1

• vesuvius.find_patches: accept -i/--input and exit non-zero when no volumes are found (revives #1631)

  #1969·mg255 opened Oct 4, 2026·1 task·

  ·
  1

  1

• pack_resident_pools: error instead of silently reusing an unmasked pool when --ct is given

  #1968·NotCqqkie opened Oct 4, 2026·1 task done·

  ·
  2