CowFIT reliable data

Artificial intelligence

The technology behind the CowFIT monitoring system was developed to collect data on individual cow’s weight and feed intake from commercial herds without disturbing cows’ behavior and daily routines on the farm.

The data from the Cattle Feed Intake System (CFIT) is used to calculate the Saved Feed index in the Nordic Total Merit index (NTM). The work on developing this patented state-of-the-art solution started in 2015.

The CFIT system has revolutionized data collection on cow weight and feed intake. Its results have been documented in several academic articles published in peer-reviewed magazines and presented at conferences worldwide.

Read the articles and conference proceedings

Journal of Dairy Science Vol. 106 No. 12, 2023

Article 1: Repeatabilities of individual measures of feed intake and body weight on in-house commercial dairy cattle using a 3-dimensional camera system

Jan Lassen, Jørn Rind Thomasen, and Søren Borchersen

The study has reported that individual feed intake and body weight can be measured using a 3D camera system that identifies the cow, predicts the body weight, and quantifies the amount of feed eaten by the individual cow. Repeatability was between 0.62 and 0.65 for daily feed intake and between 0.83 and 0.88 for body weight measured as a weekly average.

Article 2: Genetic parameters for feed intake and body weight in dairy cattle using high-throughput 3-dimensional cameras in Danish commercial farms

Coralia I. V. Manzanilla-Pech, Rasmus B. Stephansen, and Jan Lassen

The study’s promising results endorse a new technique for recording feed intake and body weight that is already in use on commercial farms. The findings demonstrate that there is moderate heritability and genetic variation in DMI and body weight phenotypes measured by 3D cameras. Moreover, estimated heritabilities and genetic correlations for both traits are similar to those reported using conventional measurement methods.

Downloads

Article 1

Proceedings of World Congress on Genetics Applied to Livestock Production

2022: CFIT – Cattle Feed InTake – a 3D camera based system to measure individual feed intake and predict body weight in commercial farms
Jan Lassen, Jørn Rind Thomasen, and Søren Borchersen

The study has concluded that feed intake and body weight can be predicted based on image analysis using a 3D camera system. The squared correlation between feed intake measured with scales and cameras was 0.90. Repeatability was between 0.86 and 0.90 for body weight.

2018: Individual cow identification in a commercial herd using 3D camera technology
Jørn Rind Thomasen, Jan Lassen, Glenn Gunnar Brink Nielsen, Claus Borggard, Peter René Bolvig Stentebjerg, Rikke Hjort Hansen, Niels Worsøe Hansen, Søren Borchersen

The study has presented a system for identification of cows based on the geometry of their back region in a commercial heard. The validation of the system achieved a precision of 90%.

How does CowFIT work?

The 3D camera and the ear tag reader are installed in the area where cows pass by one by one on their way in or out of the milking area. The ear tag is scanned to register the cow’s ID, and the camera takes pictures of the cow’s back to record its distinct body shape.

Using deep learning and artificial intelligence, each cow’s weight is identified from pictures of its back. The data from all the cows is recorded and stored in the cloud.

The results are presented in flexible dashboards and reports available for you online.

The installation is tailored to fit your barn’s conditions. One of our technicians will assess your barn and advise on your options, install the system, and perform all the necessary tests to ensure its operation.

CowFIT 3D camera takes images of the cow and cow’s ID is being registered
Images are processed with the help of deep learning (artificial intelligence)
The data from all the individual cows is recorded and stored in the cloud
The results are available for you online presented in flexible dashboards and reports

Above 90% reliability on cow weight data

The CowFIT monitoring system has been thoroughly tested to validate data and ensure the highest possible reliability.

In selected herds, scales were placed below the CowFIT camera, so images and scale measurements of the cows were performed simultaneously. This data was used to develop breed-specific, AI-based prediction models that consider all data in the image into account.

The prediction models are developed based on 21,000 images for Holstein, 19,000 for Jersey and 7,500 for Red Dairy Cattle. The squared correlation (r2) between the scale measurement and CowFIT weight prediction was above 0.90 on the visit level for all three models.

Farmers prefer not to have scales as a fixed installation, so the following verification of the CowFIT weight data has been performed based on comparison between on-farm image-based weight prediction and slaughter results, including slaughter weight and live weight (when available) on the same cows.

The data shows that there is a high correlation between the measurements: 92-94% for live weight data and 90-91% for slaughter weight data, depending on the breed.
Take a look at the detailed results.

Not all cows get a live weight at the slaughterhouse. Some places do not measure this, and sometimes, the cow does not provide a meaningful record and, therefore, has no measure.

When comparing live weight, slaughter weight, and CowFIT weight, the results are always close, and the correlation is higher than 0.90. This means that the camera’s measurements and what is observed at the slaughterhouse directly relate to each other.

	Correlation between live weight and CowFIT weight	Correlation between slaughter weight and CowFIT weight
Holstein	0.94 (671 cows)	0.91 (782 cows)
Jersey	0.92 (543 cows)	0.90 (586 cows)
Red Dairy Cattle	0.93 (478 cows)	0.91 (560 cows)

The CowFIT weight is generally a little higher than the live weight at the slaughterhouse. This also makes sense since the cow can only lose weight between leaving the herd and entering the slaughterhouse.

The CowFIT weight used in this analysis is the mean of all measurements taken in the seven days before slaughter. This also results in a lower standard deviation than the measurement taken at the slaughterhouse, which is only a single data point.

We update this analysis continuously to confirm these results as more cows with CowFIT data are slaughtered.

Easy access to data – your key to success

With easy-to-read dashboards and reports, you can access your herd’s data to get all the information you need to make well-informed decisions for your dairy business.

CowFIT empowers you with valuable insights to address nutrition, health, and reproduction problems in your herd. With reliable real-time data you can identify the pitfalls in your management routines and react promptly to prevent problems.

Rely on CowFIT data to optimise herd management in cooperation with nutritionists, vets, and advisors.

With user-friendly dashboards and reports, you can conveniently access your herd’s data on a computer or tablet to get all the information you need.

All the data is stored in one place, allowing you to track how your decisions to adjust reproduction, health, and feeding routines impact your herd’s performance.