Key points
- Our scientists have built software that uses AI to monitor cattle and their behaviours while in the holding pens of the Casino Food Co-op facility.
- The software can detect and quantify herd behaviours to determine if they are relaxed, hydrated, or displaying characteristics associated with possible injury.
- This provides evidence-based tools for animal welfare assessment, supports decision-making for best management practices, and ensures compliance with welfare standards.
Every day, thousands of cattle pass through the yards at the Casino Food Co-op in northern NSW.
Ensuring the well-being of these animals is not just good business, it is required by law. Currently, staff and vets monitor animal health and welfare. However, monitoring every animal, in every pen, is an extremely difficult task.
Can we use technology to monitor individual animals to ensure their welfare is optimised? Can technology identify if there is an issue that requires action from management? And can it ensure that animals are adequately hydrated, comfortable and healthy?
Our animal welfare and behaviour analysis specialists have teamed up with our data researchers to work with industry partner Casino Food Co-op. Together they are developing an artificial intelligence-powered monitoring system to keep an eye on animals.
How now brown cow
Monitoring animal welfare after road transport, particularly when cattle are unloaded into cattle yards, is a difficult task.
Many yards often house thousands of animals each. Even for highly trained staff and vets, continuous monitoring of every individual is an unrealistic goal. However, what if an AI-equipped camera system could keep a constant eye on them?
Identifying cattle behaviours can be difficult for a computer. Think about it: cattle of the same breed can look very similar. They have similar shapes, similar coloured hair, and stick together in groups.
So, using the power of AI, our researchers are developing software that can automatically identify cattle herd behaviours. The software can help determine if they are relaxed, hydrated, or showing signs of lameness. Using a series of cameras positioned above a pen, the technology can automatically produce video-based metrics that:
- detect animal head count
- identify animal behaviours such as lying down, standing, walking, or drinking
- estimate animal activity level
- detect lameness.
These metrics support on-site vets to identify at-risk animals and make more timely decisions on their management. It also provides objective evidence to management on how to improve their practices and ensure compliance with current industry animal welfare standards.
Like a bull at a gate
Our scientists installed cameras that continuously record cattle in the Casino Food Coop’s holding yards. Our AI-powered software then analyses the captured footage to automatically monitor how the animals are behaving.
Based on state-of-the-art AI models, our researchers developed smart algorithms to analyse animal behaviour from multiple viewpoints. This provides a robust assessment of individual animal behaviour over time. It can be used to inform on best practice animal welfare, compliance and assurance for the red meat industry.
This stage of the project has been designed to prove the technology can be useful for the industry in real-world conditions. From 2026, video surveillance systems will be mandated at Australian Meat Industry Council-certified processors under the Australian Livestock Processing Industry Animal Welfare Certification System.
"AI-powered monitoring could be a valuable addition to existing methods for ensuring animal welfare," said Dr Caroline Lee, a Senior Principal Research Scientist with us.
“Measuring animal behaviour manually at processing is challenging due to time and labour constraints. The development of a tool to automate animal welfare assessment and provide information to decision makers has the potential to transform the industry,” she said.
"Automating inspections will help the industry improve animal welfare," added Dr Dadong Wang, Research Group Leader.
“The technologies CSIRO's Data61 is developing will help make sense of the videos captured at the processing facility, provide objective evidence of animal welfare assurance, and support better decision making for management,” he said.
The team is looking to expand from tracking animals in holding pens to unloading ramps to improve early detection capabilities.
The project is jointly funded by our AI for Mission program and Data61, and supported by our industry partner, Casino Food Co-op.