the infinitely hackable annotation framework
This release include improvements based on review feedback from our JOSS submission. We want to thank our reviewers Carl Simon Adorf, Matthew Feickert and our editor Daniel S. Katz for their valuable contributions.
Add API Reference to doc pages. This release also contains some minor bug fixes.
tqdm
dependency.fit_canvas
option to adjust images to the size of the canvas by Ítalo Epifânio.This release adds a large range of usability and code improvements.
A large fraction of annotation code has been rewritten providing the basis for a range of directly user impacting improvements. The use a ipyannotator as standalone app has been simplified by updating the voila dependency. The overall code quality has been improved by becoming flak8 compliant and the gradual adding of typing checked by mypi.
This release adds video annotation support. This means you can now explore and create multi object tracking datasets right in your jupyter notebook. Many other features have been included to make this possible (bbox labeling, mulit-bbox support, and many more). Take a look at the following jupyter notebook
nbs/01d_tutorial_video_annotator.ipynb
to explore our new video annotation support.
This release provides a comprehensive refactoring of ipyannotator that allows for much better decoupling of individual annotation components. The new event driven architecture is based on the awesome PyPubSub library. The following jupyter notebook
nbs/11-build-annotator-tutorial.ipynb
provides a tutorial that demonstrates how to build new annotators based on the new architecture.
This release adds support and tutorials for explore, create and improve image classification and bounding box annotation.
01b_tutorial_image_classification.ipynb
01c_tutorial_bbox.ipynb