redpatch

A package to find disease lesions in plant leaf images

Prerequisites

Relies on Shapely.

conda config --add channels conda-forge

conda install shapely

Installation

pip install redpatch

Running in Interactive Mode

Examination of single images can be done in interactive mode. Walkthrough examples are provided in Jupyter notebooks. To start these use redpatch-start on the command line.

Running in Batch Mode

Batch processing is done on a folder of images. The script batch-process. is used to run the process. It needs three pieces of information to run.

1. A source folder - the folder to read images from, must contain images and nothing else
2. A destination folder - the folder the script will write results to. If it doesn't exist at run time it will be created. Existing folders may have their contents overwritten
3. A filter settings file - a YAML format file specifying the HSV values used in each segmentaion step. A default one can be created by the script.

Creating the filter settings file

A default filter settings file can be generated as follows:

redpatch-batch-process --create_default_filter ~/Desktop/default_filter.yml

The settings will be written to the specified file.

Analysing a folder with images with no scale card

The basic call for the basic case is:

redpatch-batch-process ---source_folder ~/Desktop/input_images --destination_folder ~/Desktop/test_out --filter_settings ~/Desktop/default_filter.yml

The script will run and output will be produced in the destination folder. The same call works whether the folder contains one or many images and if the images contain one or many leaves,

Analysing a folder with images with a scale card

If all the images contain a scale card, and the _same_ scale card, then you can get information about area added to the output. Use the --scale_card_side_length option to give the size of the scale card in centimetres.

redpatch-batch-process --scale_card_side_length 5 --source_folder ~/Desktop/input_images --destination_folder ~/Desktop/test_out --filter_settings ~/Desktop/default_filter.yml

Set the filter values for lesion centre objects

``redpatch-batch-process --source_folder ~/Desktop/input_images --destination_folder ~/Desktop/test_out --filter_settings ~/Desktop/default_filter.yml --scale_card_side_length --min_lesion_area 2 --max_lc_ratio 2 --min_lc_size 1 --lc_prop_across_parent 0.1

Analysing a folder with images with a known scale in pixels per centimetre

If you already know the scale length in pixels per centimetre you can pass that as a value

redpatch-batch-process --pixels_per_cm 472 ---source_folder ~/Desktop/input_images --destination_folder ~/Desktop/test_out --filter_settings ~/Desktop/default_filter.yml