Bring your footage
Pull clips and frames directly from cameras connected to Aura, or upload your own.
ATOM
ATOM gets your data ready for the model only you need
Annotation for detection, segmentation and tracking — fast, precise and connected straight to training on Aura.
ATOM · annotation and trainingWhy it exists
Every site has something no off-the-shelf model has seen — and good models start with well-labelled data.
ATOM is where your footage becomes training data. Label quickly, check quality as you go, and hand the dataset to training without exporting a single file.
How it works
From raw footage to a deployed model
Label precisely
Boxes, polygons, masks and tracks, with assisted labelling to speed up repetitive work.
Train and deploy
Send the dataset to training and publish the model as an agent on Aura when it is ready.
Capabilities
What ATOM handles
Detection
Bounding boxes with class hierarchies.
Segmentation
Polygons and masks for pixel-accurate labels.
Tracking
Identities across frames for movement analytics.
Quality checks
Review queues and consistency checks on every label.
Team workflows
Assign, review and approve across annotators.
Dataset versions
Know exactly what each model was trained on.
Book a demo
Have a use case no model covers yet?
Tell us about it — we will scope the data and the model with you.
Works well with
