Load and convert timelapse CZI files to movie formats

• Generate movies function that can operate on the T or Z axis
• General purpose CZI delayed reader
• Supported output formats:
  • mov
  • avi
  • mpg
  • mpeg
  • mp4
  • mkv
  • wmv

Read and interact with a large file:

from timelapse_tools import daread

# Dask array with delayed reads for every YX plane
img = daread("my_very_large_image.czi")

Generate all scene and channel movie pairs from a file:

from timelapse_tools import generate_movies

# Generates a folder with every scene and channel pair of movies in the file
generate_movies("my_very_large_image.czi")

If you want to generate these movies in a distributed fashion, spin up a Dask scheduler. The following settings generally work pretty well for our (AICS) SLURM cluster:

from dask_jobqueue import SLURMCluster
import dask, dask.distributed

cluster = SLURMCluster(
    cores=2,
    memory="16GB",
    walltime="12:00:00",
    queue="aics_cpu_general"
)
cluster.adapt(minimum_jobs=2, maximum_jobs=40)
client = dask.distributed.Client(cluster)

From there you simply need to pass the distributed executor port to the generate_movies function:

from timelapse_tools import generate_movies

generate_movies(
    "my_very_large_image.czi",
    distributed_executor_port=cluster.scheduler_info["address"].split(":")[-1]
)

It is also recommended that whichever machine you run the scheduler on, to also set the following environment variable:

export DASK_DISTRIBUTED__SCHEDULER__WORK_STEALING="False"

More details under the "work stealing" section here.

pip install git+https://github.com/AllenCellModeling/timelapse_tools.git

Note: If you want to produce workflow visualizations at the end of conversion, you will need to install pydot + graphviz.

For full package documentation please visit AllenCellModeling.github.io/timelapse_tools.

License: Allen Institute Software License