:warning: This release requires Nextflow v22.04.0 or later :warning:

New Features

• ported pipeline code to Nextflow DSL2

Fixes and improvements

• fixed a bug that lead to invalid dimensions for input images with certain aspect ratios
• moved hosting of pretrained models and test data to Azure (thanks @ebirn)
• added CITATION.cff

Dependency updates

• dropped r-shinythemes dependency as r-shiny now has bootstrap 5 support
• added version pinning for R 4.1.3
• updated r-shiny 1.6.0 > 1.7.1
• updated tensorflow-base 2.4.1 > 2.7.1
• updated imagemagick dependency 7.0.11_12 > 7.1.0_33
• updated r-corrplot dependency 0.88 > 0.92
• updated r-jpeg dependency 0.1_8.1 > 0.1_9

New Features

• added config profile for BioHPC Genomics cluster at LRZ

Fixes and improvements

• fixed illegible text in Markdown documents when browsing Github in Dark Mode (thanks @greymonroe for pointing this out)
• changed container base to micromamba for smaller container images and faster builds
• added CI tests for conda profile on macOS and additional container engines on linux (charliecloud and podman)

Dependency updates

• removed unneeded tensorflow-estimator dependency
• updated tensorflow-base 2.0.0 > 2.4.1
• updated imagemagick dependency 7.0.10_28 > 7.0.11_12
• updated shiny dependency 1.5.0 > 1.6.0
• updated tidyverse dependency 1.3.0 > 1.3.1
• updated shinythemes dependency 1.1.2 > 1.2.0
• updated corrplot dependency 0.84 > 0.88

New Features

• added charliecloud profile (needs nextflow >= 20.12.0-edge)
• added config for LRZ coolmuc2

Fixes and improvements

• improved image visualization and plot rendering in Shiny app (see #54) thanks @dschneiderch for input
• changed container registry from docker.io to quay.io
• added publication DOI to pipeline manifest
• changed (back) download links in pipeline to fetch pretrained models from Dropbox instead of Zenodo (#53)
• fixed plotting of leaf states over time in Shiny app (when appropriate metadata are provided)

Dependency updates

• updated scikit-image dependency 0.17.2 > 0.18.1
• updated imagemagick dependency 7.0.10_23 > 7.0.10_28
• updated shinycssloaders dependency 0.3 > 1.0.0
• updated slickr dependency 0.4.9 > 0.5.0

Fixes and general changes

• updated pipeline to fetch trained models from the deposited Zenodo record instead of Dropbox
• fixed an issue where the DPP addon produced sub-par segmentation results compared to the tools implementation in Deep Plant Phenomics.
• updated shiny app to show visualizations sorted by filename

New Features

• added --ignore_label parameter to exclude a segmentation class for trait calculation.
• added --masks parameter to skip semantic segmentation and run trait extraction using user-supplied masks
• added --label_spec parameter to allow for mapping of segmentation classes to pixel values of user-supplied segmentation masks. This is a requirement for the --masks parameter now.
• added --model 'DPP' and --dpp_checkpoint to allow for custom segmentation models, trained using the Deep Plant Phenomics framework

Fixes and general changes

• fixed an issue where the pipeline would crash if the input image contains an alpha channel
• updated base.config to avoid out-of-memory issues when running with --multiscale
• added log message to show current parameter settings when starting a pipeline run
• added a more informative log message if pipeline fails on systems with insufficient memory
• updated configuration for CBE cluster
• factored out Shiny dependencies into separate container (should be easier to deploy as a hosted Shiny app now)
• added Dockerfile + Conda environment for DPP v2.1.0

Dependency updates

• new (optional) dependency Deep Plant Phenomics v2.1.0
• updated shiny dependency 1.4.0 > 1.5.0
• updated scikit-image 0.16.2 > 0.17.2
• updated imagemagick dependency 7.0.9_27 > 7.0.10_23

Changes in this release:

• disabled task error strategy for pipeline runs on local computers
• fixed an issue where the network address of the shiny application was not correctly displayed on computers running MacOS

This is the first pipeline release accompanying the preprint manuscript:

Huether P*, Schandry N*, Jandrasits K, Bezrukov I, Becker C. araDEEPopsis: From images to phenotypic traits using deep transfer learning. bioRxiv. 2020 p. 2020.04.01.018192. Available from: https://www.biorxiv.org/content/10.1101/2020.04.01.018192v1

:tada: