cellseg: Multiclass Cell SegmentationΒΆ

Version 0.1.0

  • Initialized script mode.

  • Optimized imports

  • Updated docs to reflect proper DataProcessor usage.

  • Initial tests.

  • DataProcessor now errors if image and mask/label lengths differ.

  • Dropped thresholding methods. Please use pyautocv or any other image processing package of your choice.

  • Refactored show_images to handle follow current DataProcessor logic. Fixed a bug that caused rows and column switch.

  • DataProcessor now returns a dictionary containing an image, its label, and index.

  • Reading stacked tiff images is no longer supported for now.

  • dir_type was dropped in DataProcessor. Only provide a directory to image_dir containing images and masks.

  • Added data from cytounet

  • Versioning is now automated.

  • Updated docs to show that this is a work in progress.

  • Updated docs to use explicit imports.

version 0.0.0

  • Preserve name on PyPI

  • Fixed issues with show_images showing blank images for masks (labels).

  • Fixed issues with uint16 not working with PIL.

  • DataProcessor can now transform images to a given target size.

  • Renamed DataLoader class to DataProcessor to avoid conflicts with torch.utils.data.DataLoader

  • Added initial simple CNN model with a single layer

  • Added show_images in utils.py to allow quick visualization of a given number of images from a given stack of images.

  • Implemented data loaders.

  • Conceptualized project