This is a major update. Feature names, arguments and outputs have changed, so please read “Breaking changes” before upgrading from version 0.99.x.
axDistmap(): the argument allFeatures is
replaced by featureSet = "standard" or
"all".s.perimeter is Per,
h.ent.s1 is ENT, s.area is
Area). See “Feature names” in ?axDistmap. Text
files written by version 0.99.6 (original EBImage names) are converted
automatically by axSvm() and axQnt().axDistmap(): the standard output now has 11 columns
(Area, Per, Ras, Ran, ecc, mjx, ENT, SVA, SV2, CON, cir) instead of 7.
Feature tables exported with the earlier default settings lack features
required by the built-in models; please re-run axDistmap()
on the images.svm_example_model.svm is discontinued
and replaced by two built-in models, g0.1_APA (recommended)
and g0.2_ACC. AII and DI therefore differ from those of
version 0.99.x.axSvm(): the default features and hyperparameters have
changed (previously six fixed features, C = 3, gamma = 0.1). With
model = "custom" and without nCst /
nGmm, C = 1 and gamma = 0.1 are used.axSvm() now samples the training data by default: up to
12 images per group and up to 1000 objects per image, randomly selected
with seed = 1. Training results therefore differ from
version 0.99.6. To use all data, set maxNumImg = NULL and
maxNumObj = NULL.axSvm(): degenerate_path and
intact_path accept one folder or a character vector of .txt
files. Several folders, or folders mixed with files, now stop with an
error.axQnt(): the output is written to one subfolder per
input folder (<folder>_<date>_output_files),
and FileName in the summary is the path relative to the
input folder.axDistmap() and axQnt(): new arguments
pixelSize, targetPixelSize,
autoParam and colorMode.
pixelSize argument
(recommended) or from the file metadata (OME-XML, TIFF resolution tags,
ImageJ calibration).autoParam = TRUE.colorMode = "rgb" averages the three channels of
color-camera images.axDistmap() writes
<folder>_<date>_Parameters.csv (one row per
image). When analyzing images, axQnt() writes the same kind
of file for each input folder.axSvm(): new arguments model
("g0.1_APA", "g0.2_ACC" or
"custom"), features,
targetPixelSize, maxNumImg,
maxNumObj and seed. The random number state of
the R session is restored on exit. The scale of the training data is
recorded with the model when *_Parameters.csv is found next
to the input data.axSvm(): columns that are not present in every input
file are reported and not used.axQnt(): new arguments model and
distMapImg. A model produced by axSvm() is
selected with svm_model_path, and the required predictors
are read from the model itself. Models trained with version 0.99.x can
still be used.axQnt(): when images were analyzed at a scale different
from that of the model, one warning at the end of the run reports the
number of images and the scale ratio.?AiES describing supported, not
recommended and unsupported images.axQnt(): fixed a missing binarization inversion in the
image processing for images (imprtImg = TRUE), which made
the segmentation differ from that of axDistmap().axQnt(): the default of svm_model_path is
now NULL (it was tempdir()).axQnt(): objects stored in the model file (e.g. a saved
workspace) could silently overwrite the arguments of the function. The
file is now loaded into a separate environment.axQnt(): a folder, a missing file or several paths
given as svm_model_path now stop with a clear message. The
model selection dialog accepts one file only.axQnt(): no summary file is written when no file could
be processed.axDistmap(): imgType = "jpg" now works
(converted to the "jpeg" accepted by
EBImage::writeImage()).axSvm(): fixed errors when the folder selection dialog
was canceled (R >= 4.3), when a vector of file paths was given (R
>= 4.2), and in the save dialog of tcltk.axSvm(): input files with different columns
(e.g. exported with different featureSet settings) no
longer cause an rbind() error or silently dropped
rows.tiff,
rstudioapi and utils are now in Imports.Introduction_to_AiES.