Retail
Spatial and temporal annotation for retail intelligence: boxes and attributes on frames, plus timestamped events on in-store CCTV.
Who it is for
Retail intelligence vendors and in-store computer vision teams building models for store operations and shopper analytics.
Typical problems
In-store models need two kinds of label at once: what is in each frame, such as people, baskets, phones and behaviours, and what happens over time, as timestamped events on CCTV with clear start and end points.
What we can do
Spatial boxes and attributes together with temporal event annotation, delivered in rolling batches. Both layers follow the same guidelines and review, so the frame labels and the event timeline tell the same story.
Proof
Our retail surveillance AI work for a confidential client, delivered on a private CVAT instance with two-tier review, covers more than 5,000 objects per batch.
Delivery, security and quote
Every project runs through the same method: an agreed scope and taxonomy, calibration where the project calls for it, production by trained annotators and review before anything is handed back. Read more in How We Work and Tools and Delivery.
Data handling, including access, transfer, retention and deletion, is agreed with you in writing before work starts. See Security and Compliance for how we handle data.