AI technology for measuring and weighing cows

A research group at SLU has developed a tool that uses AI technology and a mobile phone camera to estimate measurements of dairy cows. The researchers became interested in the idea after seeing a commercial from Portugal claiming that it would be possible.

– It sounded complicated, but we couldn’t get the information confirmed. Since I have worked with image analysis for a long time, I wanted to see whether the technology could work and decided to apply for funding. If successful, it could make things easier for farmers, and that is how it became a research project, says Oleksiy Guzhva, Associate Professor at the Department of Energy and Technology at SLU and lead author of the study.

Thousands of points in an image

As part of the project, the researchers collected more than 10,000 images of cows in barns, taken from different angles and distances. They then used algorithms to measure the cows.

– We thought that if cows can be measured in real life, it should also be possible to measure them in an image. After processing the images with our algorithms, we were able to identify points on the cows. There can be up to 600,000 points, and the more points there are, the better the precision and correlation with the actual measurements. After three years, a lot of coffee and quite a few swear words, we got it to work!

Making farmers’ work easier

The researchers hope that the mobile app can be used in practice by both researchers and farmers. It could also support veterinarians in estimating, for example, the amount of antibiotics a dairy cow should receive when being treated for disease.

– With this technology, farmers do not need to use measuring tapes and scales, which can be both time-consuming and potentially lead to injuries. We want the app to make difficult tasks easier and provide users with valuable information, says Oleksiy Guzhva.

The app’s measurement tool is also intended to provide a basis for determining feeding requirements, investigating health-related aspects, and offering general advice.

Useful for health and welfare indicators

– We have noticed interest from both the research community and people working in practice. It can be a simple and convenient tool for quickly and efficiently collecting data for health and welfare indicators. If many people use the app, it could also generate a large collection of useful images for future projects and, in the long term, work for other cattle, horses and calves as well.

The mobile app works on iPhones with the iOS operating system, but it has not yet been launched on the market. Oleksiy Guzhva explains that additional funding is needed to make that happen.

– We see great potential in the mobile app and have learned a lot. We hope the app will become available in the future, and until then we will take what we have learned and apply it in future projects.

The research project was funded by the Swedish Farmers’ Foundation for Agricultural Research, SLF. The researchers collaborated with farmers and a digital interaction agency with experience in app development.

Read the scientific article: "PickAMoo: LIDAR-enhanced mask R-CNN segmentation for precision weight estimation in dairy cattle using smartphone imaging" in Nature.