Edit me

Introduction

Some tasks are included in the Anchor distribution, to be easily run from the command-line with -t taskname, and are termed predefined tasks.

Listing available predefined tasks

To list the names of all available predefined tasks, type in a shell either:

  • anchor -st or
  • anchor -t (without an argument)

A user can extend the available predefined tasks, by adding BeanXML to the /config/tasks subdirectory in an Anchor distribution.

How to run a predefined task

To run a task, open a shell (e.g. Command Prompt or Powershell in Windows), change to a directory with images, and then:

commandPrompt_taskName.png

Next, consider enabling additional outputs with the -o options or refining your inputs with the -i options.

See some Quick Start - Example Commands and the following tables for guidance.

Image processing

Task Name Input Type Description
histogram images extracts histograms of the intensity values an image.
resize images scales each image to fit inside fixed dimensions, preserving aspect ratio. Optionally accepts -ps.
center images like resize but uses a common output size for all images, and centers within it. Optionally accepts -ps.

Image segmentation

Task Name Input Type Description
segment/coco images instance segmentation based on 80 MSCOCO object categories.
segment/text images finds text-regions in images.

Feature extraction

Task Name Input Type Description
feature/hog images extracts a HOG feature descriptor for all images.
feature/intensity images extracts statistics of pixel values.
feature/metadata images extracts basic metadata as features, including dimensions.

Clustering

Task Name Input Type Description
cluster/timestamp any files clusters files by timestamp (from (from EXIF, file attributes or naming).

File copying / conversion

Task Name Input Type Description
anonymize any files copies files, randomizing order and hiding the original naming.
convert images converts the file format** of input images.
copy any files copies files, preserving naming and subdirectory structure.

Montages

Task Name Input Type Description
montage images creates a montage that balances available space.
montage/balance images identical to above.
montage/table images creates a montage that imposes a table structure.
montage/slices images creates a montage of all z-slices in a 3D image.

Summarizing inputs

Task Name Input Type Description
count any files counts the number of input-files.
list any files shows a line inputName -> inputPath for each input.
list/names any files shows each input’s name on a separate line.
list/paths any files shows each input’s path on a separate line.
summarize images combines the summaries in summarizeImages and summarizePaths
summarize/images images summarizes image attributes (dimensions, bit depth etc.)
summarize/paths any files summarizes file attributes (size, patterns among the file-paths etc.)

Projections across images

Task Name Input Type Description
project/mean images creates a mean-intensity projection of all inputs.
project/meanResize images scales inputs to a constant size, then the mean projection.
project/max images creates a maximum-intensity projection of all inputs.
project/maxResize images scales inputs to constant size, then the max projection.
project/min images creates a minimum-intensity projection of all inputs.
project/minResize images scales inputs to constant size, then the mean projection.
project/standardDeviation images creates a standard-deviation projection of all inputs.
project/standardDeviationResize images scales inputs to constant size, then the above projection.