cellmap_flow.ai_annotate.geometry

Which plane of the data an AI annotation reads, and which voxels it writes.

The maths is cellmap_flow.portable.ai_annotate’s, which the browser runs too, and is re-exported here under the names the server’s callers use: plan_plane (a PlanePlan: the square field around the point, snapped to the annotation volume’s voxels and clipped to the volume and the data, and the raw read covering the same physical area), raw_box (that read on the raw’s grid, as voxel (begin, end)) and the screen layout (screen_axes, screen_shape, to_screen, from_screen): the image sent is laid out as the viewer’s panel shows the plane. What is the server’s alone is reading which plane the viewer shows: its layout and its display dimensions (depth_axis_for_view, display_dimensions_for_view), which the browser reads from its own viewer (web/src/viewer/state.ts).

Axes are z, y, x throughout, as everywhere in cellmap-flow, except in the image sent (portable/ai_annotate.py says how it is drawn).

Functions

depth_axis_from_layout(layout[, display_dimensions]) → int	The data axis (0 z, 1 y, 2 x) the panel named by layout looks along.
display_dimensions_for_view(viewer_state) → tuple	The display dimensions of a neuroglancer ViewerState, as
depth_axis_for_view(viewer_state) → int	depth_axis_from_layout for a neuroglancer ViewerState, over
plan_plane(volume, volume_shape, point_nm, depth_axis, ...) → cellmap_flow.portable.ai_annotate.PlanePlan	portable.ai_annotate.plan_plane (its docstring says what the plane
raw_box(plan, voxel_size, corner) → tuple	The raw voxels to read for plan, (begin, end) on a raw grid of

Module Contents

cellmap_flow.ai_annotate.geometry.depth_axis_from_layout(layout, display_dimensions=None) → int

The data axis (0 z, 1 y, 2 x) the panel named by layout looks along.

layout is neuroglancer’s layout as a string (“xy”, “xz-3d”…), a dict with a “type”, or the state’s layout object. display_dimensions are the names of the viewer’s display dimensions in order (z, y, x when None, as a viewer that sets none over z, y, x shows them); see _LAYOUT_DEPTH_SLOT for why they matter. Only a single-plane layout says which plane the user is looking at: neuroglancer keeps one layout for the whole viewer, so in “4panel” or “3d” the panel a click landed in is not known. Those (and anything unrecognised) get the “xy” panel’s axis, the panel neuroglancer puts first. A rotated cross-section is not followed.

cellmap_flow.ai_annotate.geometry.display_dimensions_for_view(viewer_state) → tuple

The display dimensions of a neuroglancer ViewerState, as display_names gives them.

cellmap_flow.ai_annotate.geometry.depth_axis_for_view(viewer_state) → int

depth_axis_from_layout for a neuroglancer ViewerState, over its display dimensions as they are: a panel looking along one that is not z, y or x (time) gets z.

cellmap_flow.ai_annotate.geometry.plan_plane(volume: dict, volume_shape, point_nm, depth_axis: int, crop_size_px: int, data_box_nm=None, display_dimensions=None) → cellmap_flow.portable.ai_annotate.PlanePlan

portable.ai_annotate.plan_plane (its docstring says what the plane is), its refusal raised as the AIAnnotateError (“refused”, 400) the routes answer: a point outside the volume or the data, which the user fixes by moving the view.

cellmap_flow.ai_annotate.geometry.raw_box(plan: cellmap_flow.portable.ai_annotate.PlanePlan, voxel_size, corner) → tuple

The raw voxels to read for plan, (begin, end) on a raw grid of voxel_size from corner (its voxel 0’s lower corner): portable.ai_annotate.raw_box’s box.