cave-mapper

Tools for mapping between objects and representations in the Connectome Annotation and Versionining Engine (CAVE) ecosystem.

This package is a work in progress and is mainly a place to stash functions that I find useful.

Installation

pip install cave-mapper

or

uv pip install cave-mapper

or

uv add install cave-mapper

Available Functions

• map_points: This takes points, and a root ID (likely points annotated inside an object) and maps them to voxels/supervoxel IDs that are guaranteed to be part of the root ID. This is useful for mapping points to a specific object in a 3D space.

• map_points_via_mesh: This takes points and a root ID (likely points that were annotated on the surface of that object) and maps them to voxels/supervoxel IDs that are guaranteed to be part of the root ID.

Mapping points from a mesh

For a point inside an object's mesh (for example, a point in a dendritic spine), use the mesh offset and the floor convention:

from caveclient import CAVEclient, set_session_defaults
from cave_mapper import map_points

# Optional: more retries for the chunkedgraph calls than the client default. This applies
# to every client created after it, so call it before building the client.
set_session_defaults(
    max_retries=10,
    backoff_factor=1,
    backoff_max=240,
    status_forcelist=(429, 500, 502, 503, 504),
)
client = CAVEclient("minnie65_phase3_v1")
mapping = map_points(points, root_id, client, point_offset="mesh", convention="floor")

The anchor_pt_x, anchor_pt_y and anchor_pt_z columns are the chosen voxel's minimum corner in nanometers. The CAVE materialization engine looks up the same voxel for this point. Points deep inside the mesh map more reliably than points near its surface.