data labeling is dead. long live distillation. from data to object detection endpoint in 90 seconds. link:

Data labelling is dead ** .... ** unless you are working in low lighting, poor resolution and edge cases.

true!

It’s come a long long long way though!

haha i’m building the same thing

too late haha

btw I love the flow of “multiple input bboxes or polygons to PCS prompt” to allow users to annotate concepts that don’t align with SAM3s vocabulary

I honestly like prompting with boxes more than prompting with text; it’s a lot more precise

roboflow is on fire. banger after banger

It’s so cool to work with smart people; it really multiplies what you can do

will this work on construction progress drone footage too? This could be exciting for us.

What would you like to detect?

tagging different stages of a project construction and then using it for future progress monitoring of similar type of projects. Projects like solar farms, highways, and bridges to name a few.

I love this speed of execution

Working for US company feels different.

Very solid. Cool tool, nice work @roboflow

I wonder if you tried it on Bio-imaging data, like microscopy

I did with mixed results. But I’m also not a domain expert.If you are, it would be awesome if you could try it. You can do it for free.

Cool, I will. Thanks.

I'd love to hear your opinion!

In the meantime, I see no way to download label mask without downloading the entire model with the zip files. Would be awesome to be able to just get masks from the webUI, without going AtoZ.

W

That’s so cool!

This is so cool. I can’t wait to try this

dope technique indeed

90 seconds to deploy? Insane speed

Awesome

Isn’t this SAM3 ?

You should definitely monetise this into a saas , there are lots of people like me who hate manual annotating process

you killed it :)

🍳

now try with a production system beyond 1min demos, I'll wait

Is it running sam locally or nah?

@xdownloaderbot

What if there is an edge case where it doesn’t work or isn’t quite exactly what you want? Does the gui have some features that help? I imagine interactive prompt engineering but also fine tuning? Although, I imagine fine tuning isn’t as straightforward.

Could I track my warehouse packers performance with that?

is SAM hosted locally or via API ?