cellcanvas-core

Core components for CellCanvas

Requires Python 3.11+ and Zarr 3. New annotations use Zarr format 3.

Array dimensions

• Images: CZYX (channel, depth, row, column); BCZYX with a batch axis.
• 2D images keep Z=1. Add a channel axis to legacy ZYX images.
• Labels: ZYX, shared across channels; zero is unannotated. Add one to classifier predictions before saving as labels.
• Features: DZYX, where D is feature width. Use feature_axis=-1 for ZYXD inputs. Keep feature order consistent between training and prediction.