Context Images, Saved Queries, Free Tier, Batch Queries, Delete Items from GUI, Improved PR Curve and Send Videos for Polygon Annotations
Nucleus now supports context images, which can help users better understand the specific image or crop they are currently viewing. This is especially useful, when working with close-up crops or multi-camera settings. To use context images, users can simply add a new type of custom metadata with an attachment.
Try Context Images on a public data set →
We now support saved queries in Nucleus, which enables users to keep track of specific searches they performed on a given dataset or slice. If a dataset has many classes and / or metadata, queries can quickly get complex. With saved queries, it is now easy to return to a previous set of results without memorizing or externally storing the query. Create an account to save your first query →
We have recently changed our pricing structure, which now enables all users to sign-up to Nucleus for free and use the product in an unlimited manner for datasets that are smaller than 10,000 frames. We do not restrict the number of users or features, so that users can get the full value of Nucleus for free on these starter datasets.
Learn more about our pricing →
Nucleus now supports queries and filtering for the Scale annotation batch ID and batch name. This has been a frequently requested feature by customers that use Nucleus but also label data with Scale. By querying for a specific batch in Nucleus, customers can quickly inspect and control the quality of annotations.
We added the ability to send polygon tasks from video data to annotation using Nucleus. Previously, customers with video annotation projects were restricted to bounding box tasks only.
We have added a filter option for the precision recall (PR) curve, which enables users to optionally only show matches in the PR curve that have a certain minimum or maximum intersection over union (IOU).
Nucleus now enables users to remove data items from a slice as well as from a whole dataset in the UI. While most users prefer to delete through the API, removing data in the UI can be useful to immediately do small corrections.