mask2bbox 0.0.18

Gets the bounding boxes from a mask file.

mask2bbox

For a given mask, gets the coordinates of bounding box of each element of the mask. It will also allow for more operations in the future.

Instalation

pip install mask2bbox

CLI

python isolate-cells -h

Usage

import numpy as np
from mask2bbox import BBoxes

# Create a BBoxes object
all_boxes = BBoxes.from_mask("path/to/mask.png")

# Expand the bounding boxes
all_boxes = all_boxes.expand(n=10)

# Remove the bounding boxes that are located on the edge of the image
all_boxes = all_boxes.remove_from_edge()

# Get the sides of all the bounding boxes
sides = all_boxes.get("sides")

# Filter the bounding boxes by the sides
filtered_boxes = all_boxes.filter("sides", np.greater_equal, (35, 35))

# Get the IoU matrix of all the bounding boxes
iou = filtered_boxes.iou_matrix()

# Save the overlapping pairs to
filtered_boxes.save_overlapping_pairs("path/to/save/overlapping_pairs.csv")

# Save the IOU matrix to a csv file
filtered_boxes.save_iou_matrix("path/to/save/iou_matrix.csv")   

# Plot the bounding boxes on the mask image
filtered_boxes.draw(to="image", method="matplotlib", show="False", save="path/to/save/image.png")

# Save your bounding boxes
filtered_boxes.save_csv("path/to/bounding_boxes.csv")

# Get resize factors to resize the bounding boxes to a given size
resize_factors = filtered_boxes.de(desired_ratio=0.7, size=(256, 256))

# Extract the bounding boxes as images
filtered_boxes.extract(resize_factors, size=(256, 256), output="path/to/save/images")

Version Notes

0.0.15 - Update

• Update:
  • Add new entry points to the package to allow for improved CLI usage. New CLI usage is:python isolate-cells -h.
• Add:
  • Add mask2bbox/scripts folder.

0.0.14 - Update

• Update:
  • Remove LICENSE
  • Rename BBoxes.save_iou() to BBoxes.save_overlapping_pairs() in the CLI.
  • Convert BBoxes._helper_box_drawer() to static method.
• Add:
  • Add logger to the package.
  • Add BBoxes.sample() to get a random sample of the bounding boxes.
  • Add BBoxes.save_iou_matrix() to save the IoU matrix to a csv file.
  • Add -os to the CLI to save a sample of the isolated nuclei instead of the whole set.
  • Add -log to speciy the log level in the CLI (Options are: info and debug).

0.0.13 - Update

• Update:
  • Change setup.py to pyproject.toml.
• Add:
  • Added entry points to the package to allow for CLI usage.
  • Added dynamic versioning to the package version.py also accessible from the CLI python -m mask2bbox -v.

0.0.12 - Update

• Update:
  • BBoxes.from_mask() now accepts an image parameter as an additional input, this removes the need of having to load the image separately.
  • __getitem__ magic attribute indexes based on the cell/mask id (making it base 1 instead of base 0).
• Replaced BBoxes.extract_single() and BBoxes.grab() -> with BBoxes.grab_pixels_from()
  • BBoxes.grab_pixels_from() gets the pixels from the image/mask that are inside the bounding boxes.
  • BBoxes.grab_pixels_from() makes optional the processing steps implemented in BBoxes.extract().

0.0.11 - Added functionality

• Added BBoxes.grab()
• Replaced plot_to_mask() and BBoxes.plot_to_image() -> with BBoxes.draw()
  • BBoxes.draw() allows to plot the bounding boxes on the mask or the image.
  • BBoxes.draw() allows to choose between matplotlib and numpy to generate the plot.
• Modified BBoxes.extract_single()
  • BBoxes.extract_single() adds pre-processing steps such as BBoxes.extract()

0.0.10 - Added functionality

• Added extract_single(): This method allows to extract a single bounding box form the image/mask.
• Added check to BBoxes.from_mask(): Now the method checks if the mask contains any element or not and raises an error if it doesn't.

0.0.9 - Fix bug in BBoxes.extract()

• Fix bug in BBoxes.extract() that caused the images to be saved with the wrong name amd created a new folder.

0.0.8 - Change methods, change constructor, added functionality.

• Constructor: BBox object is now constructed with the syntax BBoxes.from_mask(mask) instead of BBoxes(mask).

• Operations: BBox object operations are no longer performed in place, the affected methods are:

  • BBoxes.remove_from_edge()
  • BBoxes.expand()
  • BBoxes.filter()

• Removed methods:

  • is_area_smaller_than()
  • is_area_larger_than()
  • are_dims_smaller_than()
  • are_dims_larger_than()
  • filter_identities()

These methods used in-place operations and are no longer needed, they were substituted with.

• Renamed:

  • BBoxes.get_bbox_areas() -> BBoxes.get_areas()
  • BBoxes.get_bbox_dims() -> BBoxes.get_sizes()
  • BBoxes.get_bbox_centers() -> BBoxes.get_centers()
  • BBoxes.get_bbox_ratios() -> BBoxes.get_ratios()

• New methods:

  • BBoxes.filter(): This method allows to filter the bounding boxes by any of the attributes in the object specifying the comparison operator and threshold value. It returns a new BBoxes object with the filtered values.
  • BBoxes.get(): This method allows to get the values of the attributes in the object. It works as a simplification to BBoxes.get_areas(), BBoxes.get_centers(), BBoxes.get_sizes() and BBoxes.get_ratios().
  • BBoxes.extract(): This new method isolates the bounding boxes as images and saves them to the desired location.
  • BBoxes.calculate_resizing_factor(): This method calculates the resizing factor needed to resize the bounding boxes to a given size.
  • BBoxes._pad_crop(): Helper method to pad or crop the bounding boxes to a desired output size.
  • BBoxes.plot_to_image(): This method plots the bounding boxes to the corresponding image and saves it to the desired location.

0.0.7 - Improved functionality

Improvements in IoU calculation

0.0.6 - Added functionality

0.0.5 - Added functionality

• IoU related operations:

  • boxes.get_iou_matrix() - Gets the IoU of all the bounding boxes.
  • boxes.are_overlaping() - Returns identities where the IoU is greater than 0
  • boxes.overlaping_pais() - Returns the pairs of bounding boxes that are overlaping.
  • boxes.save_iou() - Saves the IoU matrix to a csv file.

• Plots

  • boxes.plot_to_mask() Plots the bounding boxes on the mask image.

Example of boxes.plot_to_mask()

0.0.4 - Added functionality

Now is possible to extract:

• Bounding boxes dimensions BBoxes.get_bbox_dims()
• Bounding boxes area BBoxes.get_bbox_areas()
• Bounding boxes center BBoxes.get_bbox_centers()
• Bounding boxes ratios BBoxes.get_bbox_ratios()

0.0.3 - Added functionality

Added the remove_edge_boxes method to remove the bounding boxes that are located on the edge of the image.

0.0.2 - Added the expand method to expand the bounding boxes by a given number of pixels.

Added the expand method to expand the bounding boxes by a given number of pixels.
Fixed a bug in the setup.py file that made it hard to import the package.

0.0.1 - Initial release of the package.

Initial release of the package. It allows to get the bounding boxes of a mask and save them to a csv file.