We are very excited to announce the release of AnnotationLab v2.3.0. This version redesigns the Project Setup Page and introduces the annotation support for multi-page PDF documents. As always a lot of UI/UX improvements, stack upgrades, and stabilization are shipped with it. The major highlights of this release are mentioned below.
Redesigned Setup Page
With the addition of many features on every release, the Project Setup page was getting more complex for users to scroll way down on that page. Now we have redesigned the Project Setup page by breaking the main components into multiple tabs. The new Project Setup page has 4 tabs: Project Description, Sharing, Configuration, and Training.
Setup Page split into 4 sections
Along with the redesign, the Project Setup page gets a lot of UI/UX enhancements and bug fixes.
Annotate Multi-page PDF
In the last release, Annotation Lab introduced the feature to annotate on top of the image for Licensed setup, but the multi-page PDF annotation was still missing. Now the complete flow of import, annotation, and export for multi-page PDF files are available.
Users have two options for importing a new PDF file into the Visual NER project
- Import PDF file from local storage
- Add a link to the PDF file in the file attribute
After import, the user can see a task on the task page with a file name.
On the labeling page, the user can view the PDF file with pagination so that the user can annotate the PDF one page at a time.
After completing the annotation, the user can submit a task and it is ready to be exported to JSON and the user can import this exported file into any Visual NER project.
[Note: Export in COCO format is not yet supported for PDF file]Train and Test Dataset
Now Project Owner/Manager of a project can define tasks for train and test purposes using “Train” and “Test” tags respectively. These tags are now by default available in all projects.
Training with “Split dataset” checkbox selected and 23 Train tags and 2 Test tags present.
A new checkbox has been introduced on the Training Setup Page to enable the use of defined train/test tasks. Once it is checked, the val_split ratio and task filter by tags are disabled.
Enhanced Relation Annotation
The user experience while annotating relations on the Labeling page has been improved. Now annotators do not need to go anywhere to the right or bottom to choose the desired relation(s). Relations can be selected by clicking the plus “+” sign present in the relations arrow.
Get & install it here.
Full feature set here.
Try The Generative AI Lab - No-Code Platform For Model Tuning & Validation
See in action