Issues

finalize_outputs: multiclass uint8 conversion rescales each chunk by a different factor

#1432

· axiosdevs opened

on Aug 13, 2026

PHerc1203 physical-truth label grid is chunk-padded 34 voxels past the volume extent

#1407

· DarthCeltic opened

on Aug 11, 2026

vc_render_tifxyz: default --cache-gb 16 stalls silently on smaller Docker hosts, and resume counts zero-byte TIFFs as complete

#1404

· DarthCeltic opened

on Aug 11, 2026

vc_grow_seg_from_seed silently deletes the grown surface and exits 0 when the volume is an s3:// path

#1403

· DarthCeltic opened

on Aug 11, 2026

PHerc0139: 0.22% voxel-size discrepancy between the two published scans; validated cross-scan transform available

#1381

· 7jycwjmbfn-eng opened

on Aug 10, 2026

spiral-input PHercParis4: `outer_shell/meta.json` stores the grid step (~20) in `scale` instead of its reciprocal, and `vc_flatten` OOMs on it trying to allocate 1.3 TB

#1379

· Nicodol opened

on Aug 9, 2026

PHerc0500P2/...wrap01_0919: two published predictions share no top-1% pixels, alone among 56 pairs

#1377

· nerln opened

on Aug 8, 2026

Two derivations of one scan move the ink map as much as changing the checkpoint (the 8-bit window from #1211 reaches an absolute clip)

#1372

· nerln opened

on Aug 7, 2026

predict normalizes nnU-Net checkpoints with instance z-score regardless of the scheme in their plans.json

#1364

· Jinhojeong opened

on Aug 7, 2026

vesuvius.predict fails on zarr 3 (a supported version) and exits 0 while writing nothing

#1360

· TAUIL-Abd-Elilah opened

on Aug 7, 2026

Open Data catalog Lasagna attach always fails: prepareOpenDataLasagna validates before writing the lasagna-remote.json marker that validation needs to resolve remote channels

#1355

· Bullo27 opened

on Aug 6, 2026

vc_render_tifxyz: rendering at an absent pyramid level produces an all-black image with exit 0 — reachable via the sparse volumes the Las-volume flow now attaches

#1343

· Bullo27 opened

on Aug 6, 2026

Navigation Menu

Uh oh!

Uh oh!

All issues

finalize_outputs: multiclass uint8 conversion rescales each chunk by a different factor

PHerc1203 physical-truth label grid is chunk-padded 34 voxels past the volume extent

vc_render_tifxyz: default --cache-gb 16 stalls silently on smaller Docker hosts, and resume counts zero-byte TIFFs as complete

vc_grow_seg_from_seed silently deletes the grown surface and exits 0 when the volume is an s3:// path

PHerc0139: 0.22% voxel-size discrepancy between the two published scans; validated cross-scan transform available

spiral-input PHercParis4: `outer_shell/meta.json` stores the grid step (~20) in `scale` instead of its reciprocal, and `vc_flatten` OOMs on it trying to allocate 1.3 TB

PHerc0500P2/...wrap01_0919: two published predictions share no top-1% pixels, alone among 56 pairs

Two derivations of one scan move the ink map as much as changing the checkpoint (the 8-bit window from #1211 reaches an absolute clip)

predict normalizes nnU-Net checkpoints with instance z-score regardless of the scheme in their plans.json

vesuvius.predict fails on zarr 3 (a supported version) and exits 0 while writing nothing

Open Data catalog Lasagna attach always fails: prepareOpenDataLasagna validates before writing the lasagna-remote.json marker that validation needs to resolve remote channels

vc_render_tifxyz: rendering at an absent pyramid level produces an all-black image with exit 0 — reachable via the sparse volumes the Las-volume flow now attaches

Uh oh!

Uh oh!

Search results