A cashierless checkout team needed people, products and the moments between them labeled in store video, and its open-source tool couldn't keep up.
Cashierless checkout depends on understanding what people do in a store. To train it, this company needed store video labeled: people, products, and people picking products up.
They were doing it in an open-source video labeling tool that was slow, buggy and took a long time to turn data around. Just getting their own camera footage into it was overhead.
Instead of loading footage into a tool by hand, they sent it to our API. We labeled the people, the products and the moments where people interacted with products, and sent the results back.
That went well enough that a keypoint labeling project followed.
This was one of our earlier projects, from when a lot of our work was building training data for machine learning teams. It shows the same thing we do today in a different format: AI and people working through data your team doesn't have time for.
If your team is stuck labeling footage in a tool that can't keep up, this is the kind of work we did here. Talk to us about what you're working on.
