Camera-based production intelligence

PorSeen

Turn existing barn video into measurable signals for feed access, active feeding, piling, activity, and welfare risk. Operators see what happened, when it happened, and which evidence frames deserve attention.

Visible pen count
6.0 avg / 6 peak
Active feeding
214 pig-sec
Feeder body occupancy
238 pig-sec
Dense piling risk
100%
Full video analyzed
52.0 min
Active feeding
996 pig-sec
Feeder occupancy
1558 pig-sec
Detector mode
cv-fallback
Operating console

A practical command center for swine behavior intelligence.

The dashboard links annotated video, synchronized behavior trends, event jumps, and human review in one workflow. A producer can move from herd-level signal to visual evidence in seconds.

Observable behavior first

Separates body contact, head-at-feeder, active feeding, blocking, low activity, dense piling, and isolation before deriving risk indicators.

Per-pig behavior ledger

Reconciles noisy raw tracklets into six visible pig slots for individual behavior summaries.

Human review loop

Paused-frame corrections, event validation, and issue rows feed the model-improvement pipeline.

Training data path

Outputs and review artifacts are structured for model iteration and storage.

Product framing

Built to become hardware-agnostic herd intelligence.

Camera agnostic

Designed around standard IP/video feeds rather than fixed proprietary hardware.

Early warning

Risk indicators are derived from trends in observable behavior and tied back to video evidence.

Data flywheel

Corrections and validated examples become the substrate for segmentation, pose, and behavior models.