Low-cost mobile phenotyping platform to accelerate genetic improvements in structural soundness and extend the productive lives of beef and dairy cattle using 3D artificial intelligence

?Lameness is one of the top three reasons farmers must prematurely cull a cow in both the beef and dairy industries. Beyond the obvious animal welfare implications, premature loss of cattle in the production cycle is both economically and environmentally inefficient. On smaller farms, options to combat lameness through management-based interventions can be more limited, as such operations frequently have more limited access to veterinary care and hoof trimmers. Rates of lameness can be combatted, in part, by breeding for cattle with sound hoof structure, but rates of genetic improvement in hoof traits are currently limited by the accuracy of conventional visual scoring systems. OsRostrum's PhIT technology provides farms of all sizes an affordable option for next-generation phenotyping. Farmers can produce high accuracy 3D scans of key structural traits using their iPhone, producing a detailed virtual twin that they can easily review or share with others. Such scans are produced totally off-line, and when they return to a wifi connection, are sent to our cloud platform, where purpose-built computer vision algorithms process these scans and extract the same angles and proportions that our livestock judging coaches always taught us to look for. This data is then returned to farmers in individual and herd-level reports, for use in building their precision management and breeding plans, and may also be reported on to breed associations for use in national-level genetic evaluation schemes. In this project, we will collect the data and run the statistical analyses that we hope will demonstrate that adoption 3D phenotyping with realize significant improvements in measurement accuracy and reliability over visual scoring scales, that will in turn translate to improvements in trait heritability and thus faster rates of genetic gain in this key trait. We will also demonstrate the ease with which this computer vision platform can be bootstrapped to capture novel morphometric features of the hoof not currently considered in standard visual scoring scales. Finally, we will explore the feasibility of also developing a low-cost multi-camera imaging system, which will allow for faster scan acquisition in more extreme use cases, such as in hoof trimming and working with bulls.