Labelbox•June 8, 2020
We've made a number of updates to the new Image Editor, let’s dive in!
We're excited to announce a few improvements to the performance analytics available for your projects using the new Image editor:
For more details on these metrics and how to use them, please take a look at the updated documentation here.
The real-time human-in-the-loop (HIL) labeling workflow in the new Image Editor lets you feed a continuous flow of assets to your labelers for annotation, enabling a quick turnaround for your labels. It’s available upon request for Enterprise customers. For more details, take a look at our documentation page.
The model-assisted labeling feature on the new Image Editor has a new import format and improved error messaging. Still supporting everything that we had before, this new system unifies the Labelbox data model with the outputs, and allows you to close the loop quickly and easily.
A full detail of the new format can be found on the documentation page.
The new Image Editor now supports our Benchmarks and Consensus QA tools for bounding boxes, segmentation masks, and polygons.
Benchmarks and Consensus can be configured through the Settings Tab > Quality page. For more information, see our documentation for Benchmarks and Consensus
You can now use shift + s
to hide all objects on the image. You can also toggle this in the settings list by checking or unchecking “Show annotations.”
Subscribe to our YouTube channel, Labelbox Academy, to watch the latest short instructional videos of our newest labeling tools. Some of the featured tools include our new video labeling tools, superpixel, QA tools, and many others.