image-dataset-converter-imgaug 0.1.0

switched to kasperl library for base API and generic pipeline plugins

added thinning/trace-skeleton filters for skeletonizing binary images, based on Lingdong Huang’s code at https://github.com/LingDong-/skeleton-tracing/blob/master/py/trace_skeleton.py

the find-contours filter now has optional min/max size restrictions for the objects

added find-contours-cv2 filter which adds contours as polygon object detections using OpenCV

added fast-line-detection/hough-lines-prob filters for detecting lines in images and adding them as polygon object detection annotations

added simple-blob-detector filter which adds detected blobs as bounding box object detections using OpenCV

added aruco-detector filter which adds the detected markers to the meta-data

added idc-generate-aruco tool

the meta-sub-images filter can load the base filter now from a pipeline file as well

the augment_image utility method now converts numpy types back to floats for object detection bboxes and polygons

switched from imgaug3 to imaug for numpy 2 support

added roi-images filter for extracting sub-images based on object-detection bounding boxes

the find-contours filter now works on any type of image data

the transfer_region method (from _sub_image_utils) now initializes the labels of the full image segmentation annotations with the labels of the sub-image(s) to ensure correct output

fixed idc.imgaug.filter.transfer_region method: image segmentation layers now use values (0,255)

using underscores now instead of dashes in dependencies (setup.py)

renamed idc.imgaug.filter._sub_images_utils.process_image to extract_regions

added crop-to-label filter that crops the image to the bbox of the annotation with the specified label

the sub-images and meta-sub-images filter now have options to generate a regions via predefined number of rows/cols or row height/col width on the fly, useful when dealing with images of differing dimensions (and the regions parameter is now optional)

added support for placeholders: idc-combine-sub-images

idc-generate-regions tool can take margins and overlaps (right/bottom) into account for its region calculations now

idc-combine-sub-images tool now has more details in the exceptions when extraction of groups fail and prunes the annotation after the merge as well

meta-sub-images now outputs a logging message if there are no annotations after transferring regions/pruning annotations

method transfer_region now adds the sub-images rather than replacing the tile in the overall layer, to allow for overlaps (for reducing edge effects)

requiring image_dataset_converter>=0.0.5 now

switched to underscores in project name

switched to new methods from idc.api: empty_image(…) and array_to_image(…)

switched from unmaintained imgaug to imgaug3 library (https://github.com/nsetzer/imgaug)

added find-contours filter for turning blobs in image segmentation annotations into object detection polygons.

added change-grayscale filter for change pixel values of grayscale images by a factor of fixed value

added clip-grayscale filter for replacing pixel values that go below or above thresholds

added enhance filter for enhancing brightness, contrast, color, sharpness

initial release