Every fillet.
Every pellet. Accounted for.
Agavita reads a single underwater camera and turns it into live biomass, appetite and feed-waste intelligence — so bluefin tuna farmers stop feeding the seabed.
A tuna pen is a black box with a feed bill attached.
Feeding a bluefin pen is still done by eye. A farmer watches the surface, reads the boil, and guesses. Guess low and growth stalls; guess high and the excess sinks quietly to the seabed — paid for, uneaten, and gone.
Agavita puts a camera under the surface and a model behind it. The same footage a diver could barely interpret becomes a continuous read on how much fish you have, how hungry they are, and how much of today's feed actually landed in a mouth.
Segment every body in the pen
Instance segmentation separates each tuna from a crowded, back-lit, particle-filled water column — the frames a human observer would give up on. Length and girth come straight off the mask, so biomass is measured rather than guessed.
- Per-fish masks
- Length & girth
- Biomass estimate
Hold identity through the school
Multi-object tracking keeps a stable ID on each fish as it crosses, occludes and doubles back. Swimming speed, acceleration and school density become continuous signals instead of one-off snapshots.
- Persistent IDs
- Speed & acceleration
- School density
Feed to the appetite, not the schedule
Pellet detection counts what is eaten and what sinks past the school. Appetite index, feed-conversion ratio and waste land in one dashboard, so the next session is sized by evidence.
- Pellet counting
- Appetite index
- FCR & waste
Figures from the current TRL5 pilot deployment.
Put a pair of eyes in the pen.
We are onboarding pilot farms for the next deployment window. Bring a pen; we will bring the camera.