You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Data-flow audit 2026-10-29 (H1), BLOCKER. At v0.3.1.
What happens:
masks.py:387-390 requires every mask to be 2D and checks only its Y and X size.
The manifest's masks entry (:114-123) records the image, type, path and source, but no t or z.
Failure, two ways:
One 2D mask paired with a 50-frame time-lapse or a Z-stack is accepted with no plane identity. A consumer would apply it to every plane, which guesses object identity across T and Z.
Per-plane masks (TYX/ZYX) are refused as mask_not_2d, although MorphoFeature's FovFiles supports masks with T and Z axes.
Fix: refuse a 2D mask for an image with more than one Z or T plane, unless the mask carries those axes. Record each mask's planes in the manifest, and accept masks with T and Z axes.
Activity
Document mask plane identity in the README