A camera above 30 turkeys forecasts each bird's weight three weeks ahead

Days 37 to 133 of age is 96 days, a little under 14 weeks ^[1]. Mireia Molins and colleagues weighed the birds by hand five times a week across that stretch to give the network something to learn against ^[6][13]. That works out to about 68 weighing sessions per bird, and roughly 2,060 individual weights for the pen of 30 ^[2][3]. Enrico Casella, the senior author, is an assistant professor of data science for animal systems at Penn State. He said "Traditional approaches to individual body weight monitoring require extensive manual labor and frequent animal handling, creating both economic and animal welfare concerns" ^[2].

The forecast is the genuinely new part of this. Casella said earlier computer vision systems were not capable of temporal forecasting over time, and that similar forecasting had not previously been explored in the industry ^[8]. Twenty-one days is about 22 percent of the window the birds were watched over ^[4]. The system's prediction of future weight came out nearly as accurate as its estimate of current weight ^[8].

Then there is the control. All 30 birds were male, the same age, and housed together in one pen ^[3], so a predictor holding nothing but the calendar would recover some share of that 93 percent on its own. The phys.org account reports no comparison against such a baseline ^[15]. Without it, the camera's contribution above knowing a bird's age is not separable from the number.

Casella also said "In the poultry industry, weight information is essential for maximizing the value of each individual animal, as the average weight of a flock determines equipment settings for processing operations" ^[9]. That job and the per-bird job are not equally hard. Errors on individual birds partly cancel when you average thousands of them, so a flock mean is the forgiving target. The unforgiving one is also on Casella's list: identifying birds that are not growing normally ^[11].

"Imagine a commercial turkey farm with thousands of birds ... If farmers could reliably estimate weights using cameras, they wouldn't need to catch and weigh large numbers of birds manually," Casella said ^[10]. The study he is describing had 30 birds under relatively controlled conditions. He said larger studies are needed to determine how well the model works across different farms, breeds, environments, lighting conditions, poultry densities and much larger populations ^[12].