Hello there!
This is the first Label Studio Release newsletter. Since we've released the tool into the open-source around four months ago, it got great feedback from the community, valuable suggestions on where we need to focus our engineering efforts, and of course, bug reports.
To remind if you've subscribed some time ago and don't remember what our tool is about: Label Studio is a multi-type data labeling and annotation tool, it's built with the simplicity of interface in mind, and the look is fully configurable. You make the tool yourself in 10-20 lines of the config language. As a side effect of this, it is also can be embedded in your applications. Right now, it supports Images, Text, Audios, and HTML, with the goal to cover all the domains.
Now to the features of this latest (0.3.0) release and the plans for the next releases
HTML Annotation
A new annotation type! HTML is the most fundamental building block of the web, and a de facto standard to create webpages. But not only that, what makes HTML annotation powerful is how many other formats can be converted into it. For example, PDFs, now you can annotate things like resumes, research, and legal papers. Another one is excel sheets that are used for invoices and research papers. All of those can be converted into HTML and annotated. And it gives you a lot of freedom on the new data types without losing any visual representation as with text.
Still in its early days, but functional enough to be able to annotate data that comes, for example, from the excel sheets. On the animation above, you can see an invoice exported as HTML is being annotated.
Grid View - Compare annotations side by side
Often we've seen a scenario where you'd explore and compare the results created by different annotators. To make things easier, we introduce the grid view. You get all the results on one page and can compare it side by side. This way, you immediately see the differences between the annotations and explore it further.
Here you can quickly spot how different the audio segmentations are.
Coming soon
Image Segmentation with Brushes, Livewire and Floodfill
Transformers model connector for the assisted labeling
I guess this is it for now. We’d love to hear from you and get feedback on our first newsletter, what is missing, what can be improved to make it more engaging and interesting for you? Thanks for reading, enjoy your Christmas and create happy memories!
Till next time!
Michael, Label Studio team
labelstud.io