Over the past few months, we focused on helping users get started faster and be more productive with their labeling projects. Therefore, we've improved the usability of the platform. Here a the key new features.
New filtering and review capabilities
New Explore interface: Explore is now combined with the interactive query bar so that you can target the most relevant assets while reviewing
New filtering capabilities: filtering system has been revamped to facilitate search and provide additional capabilities (on categories, date, metadata etc.)
New batch actions capabilities
Batch quick actions on queue: you can manage the workflow of your assets directly from the UI by assigning, prioritizing, adding to review, and sending back to queue the selected/filtered assets
Queue UX improvements: queue navigation has been simplified so that you can easily track all your assets.
Labeling tools improvements
New job viewer: the annotation job viewer has been redesigned to show clearer information, while being faster and more intuitive to use
Shortcut customization: it is now possible to customize shortcuts for tools and classes; you can define your own shortcuts, for increased labeling efficiency
Page level classification and transcription: classification and transcription tasks can now be set to repeat on every page on a PDF document, allowing you to label every page with the same, simple ontology
New project creation workflow
Previously, to configure a project you had to do it from the Settings page. To make the process more intuitive, we designed a new multi-step project creation wizard where you can directly upload assets, set your labeling jobs, add members, and configure quality features, before entering the project space.
All project parameters, including the ones featured in the wizard as well as some secondary project-related actions, can still be edited from the Settings page.
Our Python client got a new name. Formerly known as the “playground”, it is now “Kili Python SDK”.
It comes with a new Github repo:
…and new documentation:
Behind the scenes, there are some additional changes: the SDK methods now do not require boilerplate code to handle the queries pagination anymore. This means no more having to write loops when you want to get assets or labels from a Kili project, or if you want to import the predictions from a model.
Last but not least, we are releasing the Kili CLI https://kili-technology.com/blog/introducing-kili-command-line-interface, which will let you launch operations like project creation, listing and description, asset import, and more, with a simple command from your local terminal! Check out our article for the highlights.
For additional information about changes and updates made to the platform, you can access our detailed changelog here.