macrobot package

Submodules

macrobot.bgt module

class macrobot.bgt.BgtSegmenter(image_list, path_source, destination_path, store_leaf_path, experiment, dai, file_results)

Bases: MacrobotPipeline

Macrobot analysis for blumeria graminis tritici pathogen.

NAME = 'BGT'
get_features()

Feature extraction for Bgt based on Minimum intensity projection (MinIP). doi:10.1148/rg.255055044

Returns

A list with the features per lane and it’s position sorted left to right.

Return type

list with tuple(feature, position)

get_frames(image_source)

Segment the white frame on a microtiter plate. Algorithm is based on Otsu thresholding of the UVS image.

Parameters

image_source (numpy array.) – The UVS-image (x, y, 1) which is used as source for thresholding.

Returns

The binary image after Otsu thresholding.

Return type

numpy array

get_lanes_rgb()

Calls segment_lanes_rgb to extract the RGB lanes within the white frames.

get_prediction_per_lane(plate_id, destination_path)

Predict the Bgt pathogen from the feature extraction method based on thresholding. 255 = pathogen, 0 = background

Returns

A list with the predictions per lane and it’s position sorted left to right.

Return type

list with tuple(prediction, position)

macrobot.helpers module

Image helper functions.

macrobot.helpers.get_saturation(image)

Extract the saturation channel of a 3-channel image from the HSV color space.

Parameters

image (numpy array.) – 3-channel image.

Returns

The 1-channel image saturation channel.

Return type

numpy array

macrobot.helpers.rgb_features(image_array, feature_type)

Extract min or max intensity projection.

Example Maximum RGB=[50,100,79] -> max_rgb=[0, 100, 0]

Example Minimum RGB=[50,100,79] -> min_rgb=[50, 0, 0].

Parameters

  • image_array (numpy array.) – 3-channel image.

  • feature_type (numpy array.) – minimum or maximum

Returns

The 3-channel image after intensity projection.

Return type

numpy array

macrobot.helpers.save_img_array()
macrobot.helpers.wb_helper(channel, perc=0.05)

Performing a white balance on a single channel.

Parameters

  • channel (numpy array.) – 1-channel image.

  • perc (float) – Percentile.

Returns

The channel after white balance.

Return type

numpy array

macrobot.helpers.whitebalance(image)

Performing a white balance on a 3-channel image.

Similar to GIMP white balance method.

Parameters

image (numpy array.) – 3-channel image.

Returns

The 3-channel image after white balance.

Return type

numpy array

macrobot.mb_pipeline module

Main class for the Macrobot pathogen segmentation pipelines. Version 0.4

class macrobot.mb_pipeline.MacrobotPipeline(image_list, path_source, destination_path, store_leaf_path, experiment, dai, file_results)

Bases: object

Macrobot pipeline main class.

Parameters

  • image_list (list) – A list of all images names per plate (green channel, blue channel, red channel, backlight image, UVS image.

  • path (str) – The path which contains the raw images coming from the macrobot image acquisition.

  • destination_path (str) – The path to store the final result images and csv file.

  • file_results (file object) – The CSV file for each experiments which contains the pathogen prediction per leaf.

  • experiment (str) – The experiment name.

  • dai (str) – Days after inoculation.

  • resize_scale (float) – Scale for resizing the images.

  • plate_id (str) – Plate ID without the barcode.

  • y_position (int) – Y Position for the leaves.

NAME = 'invalid'
create_folder_structure()

Create all necessary folders.

create_report()
do_whitebalance()

Calling white balance function in helpers module.

get_features()

Features extraction. Different for each pathogen should be overridden

get_lanes_binary()

Create a binary image from the lanes.

get_lanes_rgb()

Extracts the lanes of the RGB image. Different for each pathogen. Should be overwritten.

get_leaves_binary()

Segment the single leaves.

get_prediction_per_lane()

Predict pathogen. Different for each pathogen should be overridden

merge_channels()

Merging blue, red and green image to create a true 3-channel RGB image.

read_images()

Reading and resizing the images.

save_images_for_report()
start_pipeline()

Starts the Macrobot analysis pipeline.

macrobot.prediction module

macrobot.prediction.predict_leaf(predicted_image, leaf_binary_image)

Predict the pathogen per leaf.

Parameters

  • predicted_image (numpy array.) – The predicted image with the pathogen. 255 = pathogen, 0 = background

  • leaf_binary_image (numpy array.) – The binary leaf image to extract the leaf.

Returns

The infected area in percent. black_pixels * 100 / white_pixels

Return type

numpy array

macrobot.prediction.predict_max_rgb(maxrgb_image, backlight_image, rgb_image)

Predict the pathogen by thresholding the maxRGB image. Used for BGT.

Parameters

  • maxrgb_image (numpy array.) – The max RGB feature image.

  • backlight_image (numpy array.) – The backlight image used for necrosis detection.

  • rgb_image (numpy array.) – The RGB image.

Returns

The predicted image after thresholding. 255 = pathogen, 0 = background

Return type

numpy array

macrobot.prediction.predict_min_rgb(minrgb_image, backlight_image, rgb_image)

Predict the pathogen by thresholding the minRGB image. Used for BGT.

Parameters

  • minrgb_image (numpy array.) – The min RGB feature image.

  • backlight_image (numpy array.) – The backlight image used for necrosis detection.

  • rgb_image (numpy array.) – The RGB image.

Returns

The predicted image after thresholding. 255 = pathogen, 0 = background

Return type

numpy array

macrobot.prediction.predict_saturation(image_saturation, image_backlight)

Predict the pathogen by thresholding the saturation image. Used for Rust.

Parameters

  • image_saturation (numpy array.) – The saturation feature image of HSV color space.

  • image_backlight (numpy array.) – The backlight image used for necrosis detection.

Returns

The predicted image after thresholding. 255 = pathogen, 0 = background

Return type

numpy array

macrobot.puccinia module

class macrobot.puccinia.RustSegmenter(image_list, path_source, destination_path, store_leaf_path, experiment, dai, file_results)

Bases: MacrobotPipeline

Macrobot analysis for Puccinia plant pathogen. Currently works for leaf and stripe rust.

NAME = 'RUST'
get_features()

Feature extraction for Rust based on thresholding the saturation channel.

Returns

A list with the features per lane and it’s position sorted left to right.

Return type

list with tuple(feature, position)

get_frames(image_source)

Segment the white frame on a microtiter plate. Algorithm is based on Triangle thresholding of the green channel image. Different from IPK Macrobot!

Parameters

image_source (numpy array.) – The green channel image (x, y, 1) which is used as source for thresholding.

Returns

The binary image after thresholding.

Return type

numpy array

get_lanes_rgb()

Calls segment_lanes_rgb to extract the RGB lanes within the white frames.

get_prediction_per_lane(plate_id, destination_path)

Predict the Rust pathogen from the feature extraction method based on thresholding. 255 = pathogen, 0 = background

Returns

A list with the predictions per lane and it’s position sorted left to right.

Return type

list with tuple(prediction, position)

macrobot.cli module

macrobot.cli.main()

macrobot.segmentation module

macrobot.segmentation.segment_lanes_binary(lanes_roi_backlight)

Threshold the lanes by Otsu method to get a binary image for leaf segmentation. The backlight image is used for this step.

Parameters

lanes_roi_backlight (list) – The lanes of the backlight image as list of tuple(image, position).

Returns

Two list with contains the backlight lanes as binary image and it’s positions as tuple(image, position).

Return type

list

macrobot.segmentation.segment_lanes_rgb(rgb_image, image_backlight, image_tresholded, experiment, plate_id)

Extraction the lanes between the white frames. First we find and filter the contours of the threshold image to find the frames. Then we extract a rectangle inside the white frames and oder the position from left to right.

Parameters

  • rgb_image (numpy array.) – 3-channel RGB image to extract the lanes.

  • image_backlight (numpy array.) – The backlight image.

  • image_tresholded (numpy array.) – The threshold image we use as source to find the frames.

Returns

Two list with contains the RGB and backlight lanes and their positions as tuple(image, position).

Return type

list

macrobot.segmentation.segment_leaf_binary(lanes_roi_binary, lanes_roi_rgb, plate_id, leaves_per_lane, predicted_lanes, destination_path, y_position, experiment, dai, file_results, store_leaf_path, report_path, report=True)

Threshold the leaves by finding and filtering the contours of the binary lane image.

Parameters

  • lanes_roi_binary – The lanes of the binary image as list of tuple(image, position).

  • lanes_roi_rgb (list) – The lanes of the RGB image as list of tuple(image, position).

  • plate_id (str) – The plate ID.

  • leaves_per_lane (int) – maximum leaves per lane.

  • predicted_lanes (list) – The lanes of the predicted image as list of tuple(image, position).

  • destination_path (str) – The path to store the final result images and csv file.

  • y_position (int) – Y Position for the leaves.

  • experiment (str) – The experiment name.

  • dai (str) – Days after inoculation.

  • file_results (file object) – The CSV file for each experiments which contains the pathogen prediction per leaf.

Returns

Two list with contains the backlight lanes as binary image and it’s positions as tuple(image, position).

Return type

list

Module contents