scripts for dendritic data analysis and faster merging

New feature include:

• Implementation of the graph NMF method for analysis of dendritic data.
• A new notebook demo (dendritic_demo.ipynb) for dendritic data analysis.
• Faster merge iteration
• Option for merging in parallel, controlled from params.merging['parallel_merge'].
• Return correct residual during merging
• Ability for volumetric graph/sparse NMF
• Further robustness against NaNs
• Fixed some bugs for reading .sbx files #587
• More type annotations

v1.5.1

Main changes since last release:

• Some bug fixed when applying pre-computed shifts to a movie
• Updated GUI
• The params object now saves the latest github commit (if you have cloned the repo) in the data object. This could be useful for reproducibility and debugging.

CaImAn Without Borders

This release requires the creation of a new environment due to some new dependencies.

New features include:

• Support for Neurodata Without Borders (NWB) file format.
• A GUI for manual inspection and modifications of the results.
• Better support for scanbox (.sbx) files.
• Ability to save results movies and exporting results analysis.
• Ability to deconvolve traces directly within the Estimates objects (estimates.deconvolve)
• Better exporting of the results in both online and batch processing modes.
• Ability to apply all quality tests as a post-processing step during online processing.

On the backend:

• Dropping of Keras as a dependency, since it was creating issues with windows installation. It can still be used to deploy the NN models if present. Deployment is done directly with Tensorflow.
• A few additional tests for saving/loading objects.

Reproducibility note:
The metrics shown in the eLife paper for the online analysis may no longer be fully reproducible due to the new way of upsampling the results prior to saving them if downsampling was initially used. However, all the scripts still run and the results are the same.

v1.4.5

Main changes since last release:

• Speed and memory improvements in batch processing.
• Robustness against NaN values and empty patches.
• Fixed a bug arising during merging in long datasets.
• More data type annotations and better logging.

The code has been tested and works against python 3.7. However problems arise upstream in some old linux distributions. We keep python version pinned to 3.6 and will change once these issues have been resolved. For more info check #529

v1.4.4

changes since last release:

• improved dockerfile
• support for NWB files with HDF5 backend
• speedups in online processing
• type annotations and better logging (under the hood)

v1.4.3

changes since last release:

• new host for downloadable files (bypassing CloudFront)
• improved installation instructions
• minor changes and enhancements

Bug and performance fixes

Changes since last release:

• Bug fixes on recursive group save.
• Bug fix to prevent overflow when memory mapping.
• A new Dockerfile.
• Temporary pinning ipykernel to an older version to deal with slowdown in various IDEs. The pinning will be lifted once a different solution becomes available.

CaImAn forge

Main changes since previous release:

• Making installation instructions compatible with macOS Mojave.

CaImAn forge

With this release installation and upgrading of CaImAn is significantly more stable by restricting all the required packages to be installed from the conda-forge conda channel. The code will still reproduce the results of the companion paper: Giovannucci, A., Friedrich J., … & Pnevmatikakis, E. A. (2018). CaImAn: An open source tool for scalable Calcium Imaging data Analysis. eLife. In Press.

code for companion paper

This release should be considered when reproducing the results of the companion paper
:Giovannucci, A., Friedrich J., … & Pnevmatikakis, E. A. (2018). CaImAn: An open source tool for scalable Calcium Imaging data Analysis. eLife. In Press.