Magellanmapper

We have fixed many issues related to the new installed packages, including errors on first-time launch. Our first Linux build is also included!

- Added Linux package, built on Ubuntu 16.04 and tested on 16.04 and 20.04
- Fixed crash when attempting to create a log file in a new directory
- Fixed crash when the `JAVA_HOME` environment variable is not set (this setting is not required for the installed packages)
- Fixed using the bundled Java runtime environment
- Fixed updating imported image metadata when both the metadata version number and file format need to be upgraded

These packages include all of the additional updates planned for our forthcoming v1.4 release. Please see the [release notes](https://github.com/sanderslab/magellanmapper/blob/master/docs/release/release_v1.4.md) for details on these changes. One major change is the addition of logging to file, especially necessary now that MagellanMapper from installed packages is typically launched without a console. Logs are output to `out.log` located in:

- Windows: `C:\Users\<your-username>\AppData\Local\MagellanMapper`
- macOS: `/Users/<your-username>Library/Application Support/MagellanMapper`
- Linux: `/home/<your-username>/.local/share/MagellanMapper`

Any and all feedback is welcome!

Our first release with standalone installers! These packages contain a full environment without requiring use of the command-line. The Windows package includes a full installer including Start Menu integration. The macOS package contains an app bundle that can be dragged to Applications to run from the standard Launchpad.

As these packages are our first standalone releases, please do provide feedback on your install experience. More details on the changes in v1.4 in the [release notes](https://github.com/sanderslab/magellanmapper/blob/master/docs/release/release_v1.4.md).

This is a bugfix release for MagellanMapper.

Changes

Installation
- Fixed missing dependencies required by Pyface Conda dependency

GUI
- Fixed incorrect color mapping for some corresponding labels (ie same region in opposite hemispheres)

Atlas refinement
- Fixed error in compactness measurement when no surface is found

Atlas registration
- Specify a full fallback atlas profile rather than only a fallback similarity metric if the post-registration DSC falls below threshold (`metric_sim_fallback` setting)

Cell detection
- Fixed applying the first channel's profile setting for image saturation to all channels during blob detection

I/O
- Fixed reading image size and resolution metadata when values for some dimensions are missing
- Fixed importing RGB images
- Fixed redundant channel import for some formats (eg some OME-TIFF files)

Dependency Updates

- Added `import_metadata` and `import_resources` packages required by Pyface

This is a bugfix release for MagellanMapper.

Changes

Installation
- Fixed launch error when Javabridge is installed but Java is not

GUI
- The "Save Figure" button opens a file save dialog to configure the location, filename, and file format of saved figures
- Fixed display resolution on HiDPI screens set to fractional scaling (eg 150%; the fix requires Qt 5.14+)
- Fixed Atlas Editor non-Numpy image initial loading performance (regression introduced in v1.0.0, most notably in larger images)
- Fixed Atlas Editor to enable saving after edge interpolation

CLI
- Configure DPI of saved images using `--plot_labels dpi=<n>`
- Specify colors for NaN values through `--plot_labels nan_color=<color>` instead of through `--atlas_labels binary=<color>`, which is now only specifies colors when displaying labels as binary images to fix this display
- Defaults to saving images in PNG format, even if `--savefig <ext>` is not set
- CLI sub-arguments are now shown in `./run.py -h` (help documentation)
- Fixed saved plane filenames to use the plane index for the axis corresponding to the saved plane

Atlas refinement
- Smoothing metrics now include the filter size in metrics output
- Fixed the denominator in the smoothing displacement metric and smoothing metrics for non-existent labels

Python stats and plots
- Excludes decimal points that were likely included by floating-point errors
- Fixed mapping measurements to labels with weighting

This release focuses on usability fixes. New settings settings and build tools have been added to support older Linux platforms.

Changes

Installation
- `Dockerfile` based on Ubuntu 16.04 added to support running on older platforms

GUI
- Status bar shows pixel values in zoomed ROI plots
- Fixed upper/lower case in check box labels
- Fixed saving ROIs to use the original ROI offset
- Fixed resetting the labels opacity when scrolling through planes
- Fixed updating channels in the image adjustment panel for the current image type

CLI
- `--cpus <n>` command-line argument to specify the maximum number of CPUs to use for multiprocessing tasks
- ROI profile setting (`mp_max_tasks`) to set the maximum tasks per process, which can reduce memory usage considerably

Volumetric image processing
- Density heat maps default to using the whole image
- Fixed ROI saturation to use the `max_thresh_factor` ROI profile setting

I/O
- Better information when a file could not be found for import
- Specifying the `profiles` directory is no longer necessary when loading YAML profile files, including those in sub-directories
- Experiment names in the database now ignore any file extension

Dependency Updates

Python Dependency Changes

- SimpleElastix compilation
- `Dockerfile` to build for older platforms
- Allow directory to be given as a relative path in the build script
- Fixed loading a library in the script
- Install Matplotlib >= 3.3.2 (in `setup.py` but not yet in Conda install script) now that the performance regression starting in 3.3.0 has been fixed
- Additional Pandas dependencies can be installed by specifying the `pandas_plus` group (installed by default in the setup scripts)

This release streamlines refreshing and overlaying images through the GUI. Workarounds are also provided for several installation/dependency issues.

Changes

GUI
- Multiple intensity images can be selected through the GUI to overlay
- Refreshes images in each viewer after a new image has been loaded
- Selecting an ROI dropdown menu entry updates the current viewer
- Fixed retaining prior images and resolutions from previously loaded image during the session
- Fixed setting the initially displayed channel from the command-line
- Fixed to show the labels reference file if loaded from command-line
- Fixed loading ROI Editor layouts with 3D screenshots if the 3D viewer had not been opened yet
- Fixed 3D viewer orientation when opening the tab for the first time during a session
- Fixed blob detections to appear in both the ROI Editor and 3D viewer

CLI
- Simplify loading only images by registration suffix to allow specifying a directory as such: `--img <dir> --reg_suffixes [atlas-img] [annotation-img]`

I/O
- Resolutions are easier to read with fewer decimal places shown
- Filenames are saved in the database experiments table without extension to allow more naming flexibility
- Fixed applying objective magnification and zoom metadata to multi-plane image imports

Server pipelines
- Removed default microscope objective metadata now that metadata is extracted from input files when possible

Python stats and plots
- Fixed loading discrete, symmetric colormaps with an even number of colors

Dependency Updates

Python Dependency Changes

- PyQt is now specified explicitly for Conda environment installs to avoid installing an older release (ie was installing v5.9 instead of v5.12)