Penn State develops AI system to predict turkey weights with 93% accuracy, reducing manual weighing

Researchers at Penn State have created a camera and AI-based system capable of tracking individual turkeys and forecasting their body weight up to three weeks in advance, promising a non-invasive alternative to manual weighing and potential benefits for farm management.

Penn State researchers say a camera-and-AI system can track individual turkeys and forecast their body weight up to three weeks ahead with about 93% accuracy, offering a non-invasive alternative to repeated manual weighing. The work, published in Frontiers in Animal Science, points to a practical use for computer vision in a sector where even small gains in weight control can affect feed efficiency, processing settings and flock health.

According to the university, the team followed 30 male turkeys at its Poultry Education and Research Centre from day 37 to day 133, using overhead colour and depth images to capture both appearance and three-dimensional shape. The AI was trained on weights measured five times a week and used a ResNet deep-learning model to link image data with live weight.

Enrico Casella, the senior author, said traditional weighing is labour-intensive and involves handling birds, which raises cost and welfare concerns. He added that forecasting could help farms plan feed, identify birds growing abnormally and better time marketing and processing.

The researchers cautioned that the trial was small and tightly controlled. They said larger studies across different farms, breeds, lighting conditions and flock densities are needed before the method can be considered ready for broad commercial use.

Story link: https://www.psu.edu/news/research/story/camera-ai-assess-turkeys-predict-future-body-weight-study