Merge pull request #48 from Gregor-Mendel-Institute/1.2.1

merge 1.2.1 > master

CHANGELOG.md

@@ -1,5 +1,11 @@
 # Changelog
 
+## [v1.2.1](https://github.com/Gregor-Mendel-Institute/aradeepopsis/releases/tag/v1.2.1) - 2020-07-23
+
+* updated pipeline to fetch trained models from the deposited [Zenodo record](https://doi.org/10.5281/zenodo.3946321) instead of Dropbox
+* fixed an issue where the DPP addon produced sub-par segmentation results compared to the [tools](https://deep-plant-phenomics.readthedocs.io/en/latest/Tools/#vegetation-segmentation-network) implementation in Deep Plant Phenomics.
+* updated shiny app to show visualizations sorted by filename
+
 ## [v1.2](https://github.com/Gregor-Mendel-Institute/aradeepopsis/releases/tag/v1.2) - 2020-07-15
 
 * updated `shiny` dependency `1.4.0` > `1.5.0`

README.md

@@ -1,9 +1,9 @@
-# ![araDeepopsis](assets/logo/logo.png)
+# ![ARADEEPOPSIS](assets/logo/logo.png)
 
+![CI](https://github.com/Gregor-Mendel-Institute/aradeepopsis/workflows/Integration%20test/badge.svg?branch=master)
 ![Latest Release](https://img.shields.io/github/v/release/Gregor-Mendel-Institute/aradeepopsis?sort=semver)
 ![License](https://img.shields.io/github/license/Gregor-Mendel-Institute/aradeepopsis)
-[![DOI](https://zenodo.org/badge/240287274.svg)](https://zenodo.org/badge/latestdoi/240287274)  
-![CI](https://github.com/Gregor-Mendel-Institute/aradeepopsis/workflows/Integration%20test/badge.svg?branch=master)
+[![DOI](https://zenodo.org/badge/240287274.svg)](https://zenodo.org/badge/latestdoi/240287274)
 [![Nextflow](https://img.shields.io/badge/nextflow-%E2%89%A520.01.0-important.svg)](https://www.nextflow.io/)
 [![conda](https://img.shields.io/badge/install%20with-conda-brightgreen.svg)](https://conda.io/)
 
@@ -16,7 +16,7 @@
 
 # Introduction
 
-araDeepopsis is a software tool that enables plant researchers to non-invasively score plant growth, biomass accumulation and senescence from image data in a highly parallelized, high throughput, yet easy to use manner.
+`ARADEEPOPSIS` is a software tool that enables plant researchers to non-invasively score plant growth, biomass accumulation and senescence from image data in a highly parallelized, high throughput, yet easy to use manner.
 
 It is built upon the published, convolutional neural network (CNN) [DeepLabv3+](https://github.com/tensorflow/models/tree/master/research/deeplab)<sup>[[1]](#ref1)</sup> that serves the task of semantic image segmentation. A [pretrained checkpoint](http://download.tensorflow.org/models/deeplabv3_xception_2018_01_04.tar.gz) of this model has been trained upon using manually annotated top-view images of *Arabidopsis thaliana* plants of different ages.
 The code that was used for training can be found [here](https://github.com/phue/models/tree/aradeepopsis_manuscript/research/deeplab). The different models are available here [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.3946618.svg)](https://doi.org/10.5281/zenodo.3946618)

assets/shiny/app.R

@@ -26,14 +26,16 @@ library(jpeg)
 # raise file upload limit to 20MB
 options(shiny.maxRequestSize=20*1024^2)
 
-data <- read_csv("aradeepopsis_traits.csv")
+data <- read_csv("aradeepopsis_traits.csv") %>% arrange(file)
 
 labels <- Sys.getenv(c("LABELS"),"class_background=0,class_norm=1,class_senesc=2,class_antho=3") %>% 
   str_replace_all(.,"=[:digit:]+","") %>% 
   str_split(.,",") %>% 
   unlist() %>% 
   tail(-1)
 
+num_labels <- data %>% select(any_of(glue::glue("{labels}_area"))) %>% ncol()
+
 imagenames <- data %>% select(file)
 invalid <- ifelse(file.exists('invalid_images.txt'),length(read_lines('invalid_images.txt')),0)
 traitcount <- ncol(data) - 2 # exclude filename and suffix
@@ -56,7 +58,7 @@ ui <- navbarPage(title="aradeepopsis", id="nav", theme = shinytheme("flatly"),
 					tabPanel("Convex Hull",value=0,imageOutput("hull")),
 					tabPanel("Rosette",value=0,imageOutput("rosette")),
 					tabPanel("Color Channels",value=0,plotOutput("color") %>% withSpinner()),
-					tabPanel("Leaf Classification",value=0,chartJSRadarOutput("radar", height = "200") %>% withSpinner())
+					tabPanel("Leaf Classification",value=1,chartJSRadarOutput("radar", height = "200") %>% withSpinner())
 				),
 			),
 		),
@@ -109,6 +111,8 @@ ui <- navbarPage(title="aradeepopsis", id="nav", theme = shinytheme("flatly"),
 )
 
 server <- function(input, output, session) {
+		# hide radarchart if there is only one class
+		if (num_labels < 2) { hideTab(inputId = "tabset1", target = "1") }
 		# nextflow report is only generated after the run has finished once, show the tab only for resumed runs
 		if (dir.exists("www")) {
 			appendTab("nav",

conf/base.config

@@ -28,36 +28,36 @@ env {
 
 process {
   withName: build_records {
-    container = 'beckerlab/aradeepopsis-base:1.2'
+    container = 'beckerlab/aradeepopsis-base:1.2.1'
     cpus = 1
     memory = 2.GB
     time = 1.h
   }
   withName: run_predictions {
-    container = 'beckerlab/aradeepopsis-base:1.2'
+    container = 'beckerlab/aradeepopsis-base:1.2.1'
     cpus = 2
     memory = params.multiscale ? 12.GB : 6.GB
     time = 4.h
   }
   withName: run_predictions_DPP {
-    container = 'beckerlab/aradeepopsis-dpp:1.2'
+    container = 'beckerlab/aradeepopsis-dpp:1.2.1'
     cpus = 2
     memory = 2.GB
     time = 4.h
   }
   withName: extract_traits {
-    container = 'beckerlab/aradeepopsis-base:1.2'
+    container = 'beckerlab/aradeepopsis-base:1.2.1'
     cpus = 1
     memory = 2.GB
     time = 1.h
   }
   withName: draw_diagnostics {
-    container = 'beckerlab/aradeepopsis-base:1.2'
+    container = 'beckerlab/aradeepopsis-base:1.2.1'
     cpus = 1
     memory = 2.GB
     time = 1.h
   }
   withName: launch_shiny {
-    container = 'beckerlab/aradeepopsis-shiny:1.2'
+    container = 'beckerlab/aradeepopsis-shiny:1.2.1'
   }
 }

containers/addons/dpp/Dockerfile

@@ -6,4 +6,4 @@ LABEL authors="[email protected]" \
 COPY environment.yml /
 RUN apt-get update && apt-get install -y procps graphviz && apt-get clean -y
 RUN conda env create -f /environment.yml && conda clean -afy
-ENV PATH /opt/conda/envs/aradeepopsis-dpp-v1.2/bin:$PATH
+ENV PATH /opt/conda/envs/aradeepopsis-dpp-v1.2.1/bin:$PATH

containers/addons/dpp/environment.yml

@@ -1,4 +1,4 @@
-name: aradeepopsis-dpp-v1.2
+name: aradeepopsis-dpp-v1.2.1
 channels:
   - conda-forge
   - defaults

containers/base/Dockerfile

@@ -6,4 +6,4 @@ LABEL authors="[email protected]" \
 COPY environment.yml /
 RUN apt-get update && apt-get install -y procps graphviz && apt-get clean -y
 RUN conda env create -f /environment.yml && conda clean -afy
-ENV PATH /opt/conda/envs/aradeepopsis-base-v1.2/bin:$PATH
+ENV PATH /opt/conda/envs/aradeepopsis-base-v1.2.1/bin:$PATH

containers/base/environment.yml

@@ -1,7 +1,6 @@
-name: aradeepopsis-base-v1.2
+name: aradeepopsis-base-v1.2.1
 channels:
   - conda-forge
-  - r
   - defaults
 dependencies:
   - anaconda::tensorflow-base=2.0.0

containers/shiny/Dockerfile

@@ -6,6 +6,6 @@ LABEL authors="[email protected]" \
 COPY environment.yml /
 RUN apt-get update && apt-get install -y procps graphviz && apt-get clean -y
 RUN conda env create -f /environment.yml && conda clean -afy
-ENV PATH /opt/conda/envs/aradeepopsis-shiny-v1.2/bin:$PATH
+ENV PATH /opt/conda/envs/aradeepopsis-shiny-v1.2.1/bin:$PATH
 
 EXPOSE 44333

containers/shiny/environment.yml

@@ -1,4 +1,4 @@
-name: aradeepopsis-shiny-v1.2
+name: aradeepopsis-shiny-v1.2.1
 channels:
   - conda-forge
   - r

docs/parameters.md

@@ -132,6 +132,6 @@ Launch a [Shiny](https://shiny.rstudio.com/) app as the last step of the pipelin
 > R -e "shiny::runApp('app.R', port=44333)"
 >
 > # if using the container image
-> {docker|podman} run -v $(pwd):/mnt/shiny -p 44333:44333 beckerlab/aradeepopsis-shiny:1.2 R -e "shiny::runApp('/mnt/shiny/app.R', port=44333, host='0.0.0.0')"
+> {docker|podman} run -v $(pwd):/mnt/shiny -p 44333:44333 beckerlab/aradeepopsis-shiny:1.2.1 R -e "shiny::runApp('/mnt/shiny/app.R', port=44333, host='0.0.0.0')"
 > ```
 > The shiny app can then be opened in a browser by typing localhost:44333 in the address bar. It will terminate when the browser window is closed.

main.nf

@@ -110,19 +110,22 @@ Pipeline parameters
 
 switch(params.model) {
     case 'A':
-        model = params.multiscale ? 'https://www.dropbox.com/s/19eeq3yog975otz/1_class_multiscale.pb?dl=1' : 'https://www.dropbox.com/s/ejpkgnvsv9p9s5d/1_class_singlescale.pb?dl=1'
+        model = "https://zenodo.org/record/3946618/files/1_class_${params.multiscale ? 'multiscale' : 'singlescale'}.pb"
         labels = "class_background=0,class_norm=1"
         ignore_label = "None"
+        max_dimension = 602
         break
     case 'B':
-        model = params.multiscale ? 'https://www.dropbox.com/s/9m4wy990ajv7cmg/2_class_multiscale.pb?dl=1' : 'https://www.dropbox.com/s/s808kcq9jgiyko9/2_class_singlescale.pb?dl=1'
+        model = "https://zenodo.org/record/3946618/files/2_class_${params.multiscale ? 'multiscale' : 'singlescale'}.pb"
         labels = "class_background=0,class_norm=1,class_senesc=2"
         ignore_label = params.ignore_senescence ? "2" : "None"
+        max_dimension = 602
         break
     case 'C':
-        model = params.multiscale ? 'https://www.dropbox.com/s/xwnqytcf6xzdumq/3_class_multiscale.pb?dl=1' : 'https://www.dropbox.com/s/1axmww7cqor6i7x/3_class_singlescale.pb?dl=1'
+        model = "https://zenodo.org/record/3946618/files/3_class_${params.multiscale ? 'multiscale' : 'singlescale'}.pb"
         labels = "class_background=0,class_norm=1,class_senesc=2,class_antho=3"
         ignore_label = params.ignore_senescence ? "2" : "None"
+        max_dimension = 602
         break
     case 'DPP':
         model = [
@@ -133,6 +136,7 @@ switch(params.model) {
                 ]
         labels = !params.label_spec ? "class_background=0,class_norm=1" : params.label_spec
         ignore_label = !params.ignore_label ? 'None' : params.ignore_label
+        max_dimension = 256
         break
 }
 
@@ -226,7 +230,7 @@ process build_records {
                     continue
 
                 height, width = image.shape[:2]
-                max_dimension = 602
+                max_dimension = ${max_dimension}
                 ratio = 1.0
 
                 if height * width > max_dimension**2:
@@ -302,15 +306,15 @@ if (params.model == "DPP") {
 
                 for i in samples:
                     img, filename = tf.cast(samples['original'],tf.float32),  samples['filename']
+                    img = tf.image.per_image_standardization(img)
                     raw = pretrainedDPP.model.forward_pass(img, deterministic=True)
                     try:
                         while True:
                             prediction, name = pretrainedDPP.model._session.run([raw,filename])
                             logger.info("Running prediction on image %s" % name)
-                            seg = np.interp(prediction, (prediction.min(), prediction.max()), (0, 1))
-                            mask = (np.squeeze(seg) > 0.5).astype(np.uint8)
+                            mask = (np.squeeze(prediction) >= 0.5)
                             name = name[0].decode('utf-8').rsplit('.', 1)[0]
-                            imwrite(f'{name}.png', mask)
+                            imwrite(f'{name}.png', mask.astype(np.uint8))
                     except tf.errors.OutOfRangeError:
                         pass
             """

nextflow.config

@@ -77,7 +77,7 @@ manifest {
     homePage = 'https://github.com/Gregor-Mendel-Institute/aradeepopsis'
     description = "Nextflow pipeline to run semantic segmentation on plant rosette images with DeepLab V3+"
     name = 'aradeepopsis'
-    version = '1.2'
+    version = '1.2.1'
     mainScript = 'main.nf'
     nextflowVersion = '>=20.01.0'
 }