From Farm to System: Pig Genetics Now Runs on Computer Vision

For decades, choosing which animals to breed came down to two things: the technician's trained eye and a lot of patience. Visual inspection, manual measurement, and decisions made with partial data. Today, AI in the pig industry is starting to change that method at its core, with cameras and models that measure every animal the same way, every time, and turn that measurement into data you can act on.

The technician still makes the call. What changes is the information behind it.

What a Camera Can Already Measure

Automated phenotyping systems built on computer vision capture, in real time, traits that used to take weeks of follow-up: back conformation, body measurements, fat and muscle distribution, posture, and movement patterns.

The science backs it up. A study published in Scientific Reports analyzed more than 7,400 videos of 1,556 finishing pigs using automated pose estimation. Body measurements taken from video showed heritability between 61% and 74%, with a strong genetic correlation to average daily gain. These are selectable traits, measured without touching the animal.

The industry is already putting them to work. As porciNews outlines in its analysis of AI and genetics, models that estimate intramuscular fat from ultrasound images, contactless weighing with 3D cameras, and early evaluation of breeding stock are already part of the work at some of the largest integrators.

The Model Predicts, the Technician Interprets

An algorithm trained on images of animals and their production history learns which traits point to better performance. That helps identify which animals are worth breeding.

Expert judgment remains essential. The technician is the one who knows the herd's health context, walks the barns, and understands what makes each farm different. What they gain is consistently measured information, free from the variability of manual assessment. The result is selection that is more consistent, faster, and traceable: there is a record of why each animal was chosen.

The Same Camera Catches Early Signals

The impact goes beyond genetics. The same systems that identify selection traits also detect changes in behavior or body condition linked to disease or welfare issues before they become clinically visible.

Routine checks happen when the calendar allows. Early alerts arrive when the data calls for them, in real time.

A Competitive Decision

In Argentina's pig production, where margins are tight and reproductive efficiency drives much of a company's profitability, adopting these tools is no longer an innovation bet. It is how producers stay competitive.

The topic is already on the industry's agenda. In May 2026, at the 4th Argentine Pig Production Business Meeting (EEPA), organized by 333 at the Rosario Board of Trade, Juanca Santiago, CEO & Founder of Santex, gave a talk on AI and leadership to producers and executives from across the sector. His advice was straightforward:

"Don't rush out to buy technology. Invest a little and test it. If it works, scale it."

From Farm to System

The question leading producers are asking is no longer whether to bring AI into genetic improvement. It's a different one, and it has two parts:

• How much historical data does my herd have to train models on its own characteristics?

• What infrastructure do I need so information flows from the farm to the system in real time?

That is the work we do at Santex: integrating data, systems, and processes so technology reaches the point of decision. In agribusiness, we did it for a leading Argentine grain and oilseed company, developing a georeferenced traceability platform that tracks every shipment back to its source field. The result: 100% digital traceability, from concept to production in three months.

On a farm, the starting point is the same: organize the data, connect it, and get it to the person making the decision. Our AI consulting team supports that journey from the initial assessment onward.

If you want to explore which decisions on your farm could be backed by the data you already generate, let's talk about where to start.