AquaVyon
Cognitive fish counting

Count every fish, automatically

Accurate counts underpin every downstream decision, stocking density, feed volumes, survival, and harvest planning. AquaVyon counts fish and shrimp from camera feeds with computer vision, so your numbers reconcile from stocking to sale.

Up to 30%

Typical error rate in manual fish/shrimp counts

~35 min

To hand-count 2,000 post-larval shrimp

0

Fish handled, camera-based, fully non-invasive

Counting in action

See it counting, frame by frame

Every box is model output on real footage. Watch the count climb as fish move through the frame, including the ones at depth and partly hidden behind others.

Try it on your own footage
Live fish counting in a dense school
Detection and tracking running frame by frame, with a live count and near/mid/far distance breakdown.

Why manual counting costs you

Hand-counting is slow, stressful for the stock, and error-prone at scale, and a wrong count early cascades through the entire crop. Vision-based counting is fast, non-invasive, and consistent.

  • Counts from camera feeds, even in dense schools
  • Works in turbid RAS, pond and pen water
  • Consistent numbers across the whole cycle
  • From post-larvae shrimp to grow-out finfish

Stocking

Verify counts at stocking so density, feed and forecasts start from a real number.

Transfers & grading

Reconcile numbers across moves and gradings without hand-counting.

Harvest

Confirm counts at harvest and reconcile inventory from stocking to sale.

Trials

Accurate survival and count data for feed, vaccine and stocking trials.

How vision-based counting works

1

Camera captures fish or shrimp passing naturally

A fixed or handheld camera observes stock moving through a chute, tank, or transfer point, no traps, no confinement, no handling.

2

Vision model detects and tracks each individual

The model is trained to separate overlapping bodies in dense schools and low-visibility water, so it doesn't double-count or miss fish hidden behind others.

3

Counts reconcile against expected numbers

Running totals are checked against stocking and prior-count records, flagging discrepancies immediately instead of at the next manual audit.

Fish counting FAQ

How accurate is AI fish counting compared to manual counting?

Manual fish and shrimp counting commonly runs up to 30% error, especially at volume, a vision-based count from camera footage is consistent across the whole batch rather than an extrapolated hand-tally, and we benchmark against known stocking numbers on your farm during onboarding.

Does AI counting work for shrimp post-larvae, not only finfish?

Yes, counting post-larval shrimp is one of the most time-consuming manual tasks in a hatchery (roughly 35 minutes to hand-count 2,000 shrimp). See the dedicated Shrimp Intelligence page for shrimp-specific detail.

Can it count fish in dense schools or turbid water?

Yes. The models are trained to separate overlapping bodies in dense schools and to work in turbid RAS, pond and pen water, not only clear, sparse conditions.

Do I need to install a camera to try fish counting?

No, you can send a video of your fish or shrimp and get a free automated count back in 96 hours, with no hardware install required.

Counting fish or shrimp? Send us a clip.

Get a free automated count from your own footage in 96 hours, no install required.

Free count in 96 hours