STTR Phase I: Swine Automatic Lameness Sensor (SALS)

Motion Grazer Ai, Inc., East Lansing MI

Investigators

Abstract

The broader impact/commercial potential of this Small Business Technology Transfer (STTR) Phase 1 project is to provide an in-farm sensing system that will notify sow (adult female swine) farmers of early signs of animal lameness, and thereby reduce early sow mortality and enhance farm productivity. The technology uses artificial intelligence to analyze pig locomotion in order to spot subtle patterns indicative of lameness. Early detection of lameness will enable farmers to take corrective actions rather than waiting for lameness to deteriorate to sow death or culling. Early culling or sow death is a major economic cost to farmers and a large fraction of death and culls is due to animal lameness. Successful application of the technology being developed in this project promises to reduce early sow mortality and culling, leading to additional litters per sow and so provide a significant economic boost to farmers. With patent applications for key components of the sensing system, farmers will install sensors in hallways and obtain health measures for each sow when she moves between rooms. The projected annual revenue is $3.0 million. This Small Business Technology Transfer (STTR) Phase I project proposes combining an imaging sensor with artificial intelligence to create a unique sensing system to unobtrusively and remotely diagnose lameness in sows (adult female swine) as they traverse hallways. This project seeks to validate two key technical contributions. First, precise 3D animal posture and locomotion are estimated for sows moving beneath a ceiling-mounted sensor. High accuracy is achieved through a novel annotation technique that overcomes difficulties in inaccurate manual location of skeletal landmarks. Second, a data-driven approach is used to train a deep neural network to learn the most discriminating combinations of posture and gait for determining lameness in walking sows. A self-supervised neural network sidesteps the need for extensive manual annotation and expert annotation is only required for lameness assessment. Together, these two contributions will enable a transformative technical capability of a remote sensor that can automatically diagnose early-stage lameness in sows. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.