New Sidebar, Billing Page, Free Credits, Custom Object Embeddings, Sort by Scene Metadata, Delete Items From Slices, Ego Distance Metric, 3D Precision-Recall Curve


🧭 Simpler navigation with a new sidebar

The new Nucleus sidebar makes it easier to navigate within your dataset. Besides direct access to Charts and Autotags associated with a dataset, you can now search and filter your slices directly in the sidebar, which makes it faster to find your previous selections.
Explore your data with the new sidebar β†’


πŸ’³ New billing page and $200 in free credits

The new billing page enables you to review and edit all payment related information directly in Nucleus. Update your payment method, review past invoices, add members to your team or change your subscription status. We are happy to share that every new user receives $200 in free usage credits on Nucleus.
Claim your free Nucleus credits today β†’


🧊 Custom object embeddings

Nucleus now allows you to upload custom embeddings not just on the image level but also on the object level. This enables you to use our powerful similarity search and Autotag features on objects based on your own model embeddings, making data search and curation more fine-tuned for each use case.
Learn more about custom object embeddings β†’


What else is new?

πŸ“‚ Sort by scene metadata

You can now sort query results based on metadata which is at the scene level. This complements metadata searching and sorting on image and object level and further enriches the ability to identify the highest value data for your model from sequences.
Learn more about metadata in Nucleus β†’

❎ Delete items from slices

Nucleus now allows the deletion of items from a slice via the GUI. This makes slices more versatile and flexible when using them for queries, model performance analysis or sending them to labeling.
Learn more about slices β†’

🚘 Ego distance metric

For scene datasets, Nucleus now enables filtering on ground truth and model predictions based on their distance to the ego vehicle recording the data. This is a powerful feature for autonomous driving teams to understand their perception model performance relative to the distance of the detected objects.

πŸ“‰ 3D precision-recall curve

The precision recall curve (PR curve) on the Nucleus charts page now also supports 3D sequential datasets, which enriches the analytical capabilities of Nucleus on these datasets.