Configuring NBLAST

After setting up the R environment for NBLAST the front-end tools like the Similarity Widget should work. By default these tools will read the most recent skeleton data from the database. For bigger datasets this can introduce some performance problems. Typically, in bigger datasets, only a small portion of skeleton does actually change and a cache can be used for some data.

Creating skeleton caches

Caches are stored in R’s binary RDS files in the cache directory in the MEDIA_ROOT path. At the moment, caches are created either manually, e.g. through the management shell or through a cron job:

from catmaid.control.nat.r import create_dps_data_cache
project_id = 1
create_dps_data_cache(project_id, 'skeleton', tangent_neighbors=5, detail=10, min_nodes=100, progress=True)

This would create the cache file r-dps-cache-project-1-skeleton-10.rda`, following the pattern r-dps-cache-project-<project-id>-<type>-<detail>.rda.

Caches can be created for the types skeleton and pointcloud. The tangend_neighbords settings defines how many neighbor points should be used to compute a tangent vector, the default is 20, but 5 often yields good results as well and is faster. The detail setting defins the branching level below which skeletons will be pruned. Using the min_nodes setting, only skeletons with the respective minimum number of nodes are included. By default, no progress is shown, which can be changed using the progress setting.