Research program

The science behind behavior intelligence for swine production.

PorSeen turns ordinary barn video into validated behavior signals: feeding access, active feeding, rest, piling, activity, isolation, and competition. Those signals become early-warning workflows producers can act on before performance or welfare problems compound.

Vision ethogram

Behavior signals captured from camera feeds

Review evidence
Counting and visibility
Feeder use
Drinker use
Posture and rest
Activity level
Dense piling and thermal comfort
Social proximity and isolation
Competition and displacement
Aggression and injury risk
Growth and body condition

Observable labels

The vision stack focuses on what a camera can verify: visible count, body contact, head-at-feeder, active feeding, lying/resting, piling, low activity, isolation, and competition.

Risk indicators

Longitudinal behavior trends surface feeding-drop risk, low-activity risk, dense-piling risk, and resource-competition risk in a form producers can inspect and review.

Outcome research

As outcome labels accumulate, the same evidence pipeline can connect behavior changes to growth, lameness, respiratory, and feed-conversion outcomes.

Data flywheel

Every reviewed frame sharpens the system.

Capture

Raw videos and generated annotated replays are retained as evidence.

Review

Humans label paused frames, event samples, and issue rows.

Train

Reviewed frames become segmentation and pose candidates.

Validate

Count, feeding, piling, activity, and event metrics are scored after each iteration.