When I first started building with visual AI, I quickly learned that you need to develop a labelled image dataset. What is that? What do you do? How do I start?
I was attempting to use AutoGluon, and it ingested something called a "COCO dataset". I quickly looked it up; it makes sense and looks simple enough, but how do you draw the shape around the thing you're interested in? I saw videos of people doing stuff on Gimp; I saw services charging a few hundred dollars a month. They all seemed more than I wanted.
I just wanted to draw squares over images, so Annoset is precisely that. There is little more. I don't want it to be too complicated.
A couple of weeks ago, I had a spare weekend and thought, "What would it take to build something in a weekend?" I did it, but this is the more refined version.
This has no extensive feature list; it just lets you upload an image, draw shapes around boxes and get some JSON in return.
Annoset is targeted at people like me—somebody doing some R&D, just setting out, who just wants to have a play. Nothing more. If you're a large organisation, move on. I built it for myself, and I want to share it with you.
If you're looking for a simple tool to generate some COCO datasets, try it. If you have a problem, an idea, a request, etc., get in touch via the contact form.
Adam