Countop
Sports analytics. They processed every event by hand for one or two weeks. We built a computer vision system that does it on its own and leaves people only the review.
Result
less processing time
less manual review
lower cost per event
The problem
Countop delivers footwear and audience data to major sports brands. Every event generated thousands of images that somebody had to classify by hand. The bottleneck was neither commercial nor technical: volume grew faster than the capacity to review it, and that capped the business.
What we did
An automatic processing pipeline with models trained on their own images, because generic models could not tell apart the details their clients pay to know. On top, a review application where a person confirms or corrects in seconds instead of classifying from scratch.
Have a similar bottleneck?
If repetitive work on images is capping your capacity, it can probably be moved.