All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- SIA: Delete annotation by hitting Backspace
- Example pipeline for multi label support in SIA
- SIA:
- Close label input field also when clicking on a annotation
- Autoscale font size in AnnoBar
- MIA: Frontend undo button
- Heavy frontend cleanup
- MIA: Allow to assign label by mouse click in dropdown also if underlaying image is excluded
- anno_helper.divide_into_patches: output of shifted bbox coordinates.
- blacklist: fixed behaviour of get_whitelist method. Do not put images automatically on blacklist.
- SIA: Added button to delete all 2D annos
- Update AnnoTask progress bar on sia update events
- adapted imageai package installation for gpu container
- Blacklist class for image blacklisting
- Added anno helper & vis to docs
- Removed LifeSign from Logger
- LifeSign from logging
- frontend cleanup: removed old tools (sia/pipeline)
- fixed destroyed conda env in lost-cv-gpu container
- SIA: New features
- Fullscreen mode
- New GUI design in order to enlarge the image canvas
- Assignment of image labels for captioning
- Multi label support -> Assignment of multiple labels per annotation or image
- Full redo/ undo support
- Introduced info boxes to show additional information
- An annotator may mark an image as Junk if it should not be considered for annotation
- A minimum area for annotations can be defined.
- Database:
- Added description field to ImageAnno and TwoDAnno. This can be a description that can be added by an annotator or algorithm und was added for future features.
- Added new table(track) for track annotation, to prepare the database for the ISA (Image Sequence Annoation) tool.
- SIA: Complete rewrite in react.
- Config of SIA in pipeline definition files.
- Database:
- In ImageAnno changed track_n -> track_id
- SIA: In pipeline definition files config -> actions -> edit. It is not longer possible to allow/ deny actions when editing an annotation, since it was a feature that nobody used.
- Cron: Set AnnoTask to finished if there no annotations in the current iteration
- Docs: Added description of @KonradAdamczyk on how to install LOST from a backup
- LOST notification system: @KonradAdamczyk fixed missing lost_url
- GPU Worker lost-cv-gpu now contains the scikit-learn library
- Annotation context can now also be added to monochrome images
- Fixed typos in pipeline gui
- Fixed import of script ENV variables that have been commented out (see commit 72fddfa)
- Fixed offline availability since frontend was loading fonts from web
- allow pipelines to be started by users other than admin (groups/user id mix-up)
- allow arbitrary ports for api access through lost frontend
- Check if pipeline definition files match specification before import or update.
- Allow custom nginx configuration.
- allow assigning annotasks to all users and groups in system (see issue #19)
- change label of predicted annotations in sia (see issue #23).
- prevent users from deleting themselves.
- prevent users from deleting their default group, which will prevent pipelines from starting.
- go to last image in sia: not one image too far anymore (see issue #31).
- removed redundant data for sia annotations (see issue #29)
- added jwt blacklist
- Adjust loop iterations when starting a pipeline
- prevent admin user from removing its own designer role
- update_pipe_project.py bug: update of wrong script (see issue #21)
- logout bug: clean redux store and reset axios auth header (see issue #18)