Sorta
Sorta AI WorkspaceHelping people — and eventually machines — tell one kind of waste from another.
The problem
Most of us stand in front of a bin every day and guess. Rules differ block by block, labels contradict each other, and good intentions end up in the wrong container. The result is recyclable material sent to landfill and contaminated loads that make recycling more expensive for everyone.
Our purpose
Sorta is building a carefully governed collection of waste imagery so that everyday waste can be understood object by object rather than bag by bag. Getting that groundwork right — clear categories, honest labels, and a review process people can trust — matters far more than moving fast.
The long-term vision
A tool that can look at a pile of waste and name what it sees, giving households, schools, and community programs plain guidance about what belongs where — and giving waste operators a clearer picture of what is actually flowing through their streams.
Eight broad categories
Everything starts from a small, stable set of material families. Narrower rules are added deliberately over time rather than invented up front.
- Organics
- Fiber
- Plastic
- Metal
- Glass
- Inert/Mineral
- Hazardous
- Residual
Where the project stands
The foundation and annotation workflow are built. Phase 3 focuses on reviewed data, versioned datasets, and a local training workflow. A trained recognition model is not yet available on this website. Progress updates live on the documentation page.
Source images are private and are never published on this site. Team members can sign in to the workspace to contribute.