We built Score Studio so you can go from raw footage to a deployed model without the dataset and labelling grind. So we're curious what you'd point it at first.
Anything goes, a defect on a production line, people in a car park, a specific object in your footage. Drop it below and we'll run the interesting ones!
Hey folks,
Max here.
We built Score Studio because the same thing killed almost every computer vision project we've ever built: collecting the footage, labelling it by hand, and training for weeks, only for the model to fall apart the moment it hit the real world. So we automated the whole path. Drop in raw footage and Studio builds the dataset, labels it, trains, evaluates, and gives you a model you can deploy, in the cloud or on your own machine. For the hard problems, teams compete on a bounty and you keep the best model.
It's free to try on your own footage. It's early and not perfect yet, so I'd genuinely love your honest take, good or rough!
If you give it a go, tell me what task you gave it and where it fell short. That's the most useful thing you can leave us today.
Hey everyone, Khalil here 👋
I've been working on getting Score Studio out into the world, so it's a good day.
The short version: you point it at footage, it handles the dataset, labelling, training and evaluation, and hands you a model you can actually deploy. No more weeks lost to labelling.
Genuinely curious what it does for you. If you try it, tell me the task and where it held up or fell over, would be happy to help!