Skip to content

Latest commit

 

History

History
79 lines (62 loc) · 4.94 KB

README.md

File metadata and controls

79 lines (62 loc) · 4.94 KB
Raw

ShinyFruit

An R shiny application for interactive image phenotyping of fruits and vegetables. The main goal of creating this app was to provide a reasonably simple and tailored solution for breeders and horticulturalists to efficiently collect data about size, color, and damage across large numbers of images. The user simply sets parameters for a single sample image and those settings are applied to all images in a given directory. In its present state, one can use ShinyFruit to measure the following traits:

  • Length
  • Width
  • Area
  • Fruit Count
  • Proportions of Discolored Regions (i.e. Red Drupelet Reversion in blackberry)
  • Median RGB values
  • RHS descriptive color profiles

Please feel free to share any feedback, bugs, or new feature ideas with me so that I can continue to improve this tool!

Installation

install.packages("devtools")
devtools::install_github("mchizk1/ShinyFruit", build_vignettes = T)

Check out the vignette for a full tutorial

vignette("ShinyFruit-Tutorial")

Or just jump right in by booting up the app!

ShinyFruit::run_app()

What's happening behind the scenes?

When the user loads up an image, it is automatically edited with the magick package in a few ways to optimize for speed and feature detection:

  • Images are downsized so that the maximum dimension is capped at 1500 pixels (for speed).
  • Contrast is increased by normalizing pixel values to span the full color range.
  • Differences in color intensity are sharpened
  • Images are enhanced to reduce noise or inconsistencies

After the image is loaded, and the user moves on to set thresholds for background removal, where images are automatically despeckled to remove small, unimportant islands of pixels. This is important because the user will not want to count all the spots of dirt or juice when it's time for analysis. This process is achieved by first shrinking and then swelling detected islands of pixels. In the end, only the larger objects will remain. Despeckling becomes optional when using the "Color-Based Feature" trait in the analysis stage.

Output data dictionary

The csv output file produced by ShinyFruit can contain a number of different variable columns, which are described in the table below:

Variable Context Description
File General The name of the file analyzed
Colorspace General Color space chosen to remove background area
BkgCh1Threshold General Value set for color channel 1 in background removal
BkgCh2Threshold General Value set for color channel 2 in background removal
BkgCh3Threshold General Value set for color channel 3 in background removal
MeanRGB Color profile Mean RGB value of the image (after background removal)
RHSDarkColor Color profile Nearest Royal Horticultural Society color value to the darkest value of the image (after background removal)
RHSMidColor Color profile Nearest Royal Horticultural Society color value to the median value of the image (after background removal)
RHSLightColor Color profile Nearest Royal Horticultural Society color value to the lightest value of the image (after background removal)
FtColorSpace Color feature analysis Color space chosen to analyze a colored feature
FtCh1Threshold Color feature analysis Value set for color channel 1 in color feature analysis
FtCh2Threshold Color feature analysis Value set for color channel 2 in color feature analysis
FtCh3Threshold Color feature analysis Value set for color channel 3 in color feature analysis
Feature_prop Color feature analysis Proportion of the the image occupied by the color feature (after background removal)
FeatureRGB Color feature analysis Median RGB value of the extracted color feature
FeatureDarkRHS Color feature analysis Nearest Royal Horticultural Society color value to the darkest value of the color feature
FeatureMidRHS Color feature analysis Nearest Royal Horticultural Society color value to the median value of the color feature
FeatureLightRHS Color feature analysis Nearest Royal Horticultural Society color value to the lightest value of the color feature
BerryCount Size analysis Number of fruit detected after removing background area
Length Size analysis Mean length of fruit detected
Width Size analysis Mean width of fruit detected
Size Size analysis Mean area of fruit detected

To print this table in R simpley run the function show_variables()