Impact
A number you can act on.
Cleanup budgets get allocated on impressions. A survey replaces the impression with a count.
Coverage, not sampling
A transect walk samples a strip and extrapolates. A drone pass records every item in the flight area.
Per 4K frame
Roughly a second per full-resolution frame, so a morning's flying is processed the same day.
Classes itemised
Counts are broken out by material, so glass and plastic are budgeted separately.
What you get back
- Annotated frames — every detection drawn on the original imagery, so any number can be traced back to the pixel it came from.
- Per-class totals — count, footprint area and estimated mass, broken out by material.
- Survey totals — the same figures for the whole flight area.
- The imagery itself — kept, so a later survey of the same stretch is directly comparable.
Repeatability is the point
A single count says how much litter is on a beach today. The same method flown again in three months says whether anything you did worked. Because the pipeline is fixed — same model, same scale calculation, same reference weights — two surveys of the same stretch differ only by what actually changed on the ground.
Corrections feed back
Where the model gets something wrong, a human fixes the box and the correction is written into the training set. The model improves on the imagery it is actually being used on, rather than on a benchmark that resembles it.
What this is not
It is a measurement instrument, not a cleanup service. It tells you where the litter is, what kind it is and roughly what it weighs. Removing it is still work for people with bags.
Bring us a coastline.
We will process a sample survey and show you the output before you commit to anything.
Request a demo