Livestock Health

How Sigma’s AI platform can estimate the cow’s body condition score.
livestock

The Challenge

Body condition can be used to early detect transition cow diseases (TCD). Currently the body condition score (BCS) is mainly scored visually and infrequently, so early detection of diseases is difficult. One of the world leading research groups of Ruminant Population Health based in the UK wanted to test the application of artificial intelligence to the body condition score estimation.

Sigma’s Computer Vision recognizes patterns with a very high accuracy, including thermal image patterns.

The Solution

The research group together with Sigma tested a number of thermal video camera locations and analyzed the image quality obtained from each location. Once the right video camera location was determined, the research group collected a large number of thermal video sequences from a population of cows and annotated the body condition score for each image using Sigma’s data annotation tools. Sigma utilized the annotated data to adapt its computer vision algorithms to estimate the body condition score.

The Outcome

The estimation of the body condition score had two steps: The cow detection and the estimation of the BCS. The cow detector accuracy was 96.5% and the BCS estimation accuracy was 83.5%.

Given the success of the first version of the BCS estimation system, there is a second phase in which more data from a larger cow population will be collected.