predictSD-v1.2.0 Filter Label Images

• Filters are now also applied on the label-containing TIFF's and output is saved to folder "filtered".
• Intensity slope calculations can be switched on/off via setting 'calculate_slopes'.

Full Changelog: v1.1.2...v1.2.0

predictSD-v1.1.2 Overlay and Slope bugfix

Improved handling of overlay plots:

• ImageJ color LUT selection based on available .lut-files
• Bit depth changes in images are now handled
• Related paths are now explicitly resolved

Improved collection of label information:

• Handling of NaN-values in data
• Intensity Slope bugfix

Full Changelog: v1.1.1...v1.1.2

predictSD-v1.1.1 Linux Compatibility

• Fixed paths to be correctly resolved.
• Overlay images can now be created using the default Linux ImageJ installation

predictSD-v1.1.0 Label Filtering

Filters can now be applied to label data of one or more models. This can be done in several ways:

1. Modification of the value of prediction_configuration keyword 'filters' before executing labelCollect.py.
2. By giving a list of filter tuples to CollectLabelData.__call__(), i.e. filters=[(model, column, value, filter type)]
3. By giving a similar list of tuples CollectLabelData.filter() after the label information has been collected.
4. Use CollectLabelData.output.filter_output() (i.e. OutputData.filter_output()) to apply a filter to a single label DataFrame.

predictSD-v1.0.2 Memory Management

Available GPU memory and fraction to dedicate for prediction can now be defined via prediction_config keyword 'memory_limit'.

• Previous versions at times caused memory errors due to faulty allocations.

predictSD-v1.0.1 Environment update

Setup instructions and predictSD dependencies updated for CUDA v11.4

• GPU calculations were disabled in previous environment (CUDA v10.1) if using StarDist v0.7.*

Publication release

Released in conjunction with LAM publication (doi.org/10.1101/2021.01.20.427422 ).