About
Measure first.
Most decisions about litter are made on numbers nobody measured. This is an attempt to fix the measuring.
Why this exists
Ask how much litter is on a stretch of coastline and you will usually get a figure derived from a transect walk — a volunteer counting items in a few sample squares, multiplied up. It is careful work, but it is a sample, it is slow, and two people doing it produce two different answers.
A drone flies the same stretch in minutes and records every frame. The problem stops being collection and becomes reading — which is a problem computers are good at.
What we care about
- Tight boxes over impressive ones. The headline detection score was already high before the biggest fix; what was wrong was box tightness, and that is what the mass estimate rides on.
- Sourced numbers. Every reference weight has a citation. Our own estimates were off by up to 59% when we checked them, which is why we stopped estimating.
- Stated uncertainty. Mass is called an estimate everywhere it appears, because that is what it is.
- Humans in the loop. Every prediction can be corrected, and corrections become training data.
IIDA
SAAF is built at IIDA, a focused research organisation. Sensing and Accounting of Aquatic Floatables is the long form of the name — it describes the two halves of the problem: sensing what is there, and accounting for how much of it there is.
Talk to us
If you run surveys, manage a coastline, or study marine litter, we would like to hear what you need out of a tool like this. Write tohello@iidalabs.org.