New start-up launched to bring computer vision tech to farms

There have been many start-ups attempting to link next-generation monitoring technology with disease detection in recent years. The newest addition to this landscape is Vet Vision AI. At the recent Animal Health, Nutrition and Technology Innovation Europe conference in London, S&P Global’s head of animal health Joseph Harvey spoke to the company about its dedication to improving health and welfare.

Vet Vision AI has been formed to commercialize machine learning technology across multiple species, initially focusing on livestock. The start-up is a spin-out from the University of Nottingham in the UK. It is the brainchild of Charles Carslake and Robert Hyde – both veterinary surgeons that moved into academia.

The company applies standard cameras to analyze on-farm animal behavior and track potential disease outbreaks. Dr Hyde believes Vet Vision AI’s technology platform “can detect disease quicker than people can and much more accurately”. However, he pointed out the business is keen to not only detect disease but also aid prevention. The latter is something Dr Hyde believes is missing from the capabilities of other disease detection systems currently being adopted by farms.

He remarked: “We’re trying to distill our decades of veterinary expertise into an algorithm that can watch what animals are doing in their environments 24 hours a day. There are lots of companies out there that say they can detect disease in animals very quickly, which is great because you can treat the animals quickly too. Then another case pops up and another, and you continue to treat but you don’t really get any closer to understanding why these cases keep happening. We want to quantify the management and behavioral indicators that drive disease.

“We want to be able to allow vets to pinpoint exactly what is going on. We think this is really crucial for disease prevention and for animal welfare too. Obviously, we still want to help treat the animal as soon as possible but we also need to identify the actual problem and help the farmer on a more long-term basis.

“Lameness is a good example of this. There are several companies detecting lameness in different ways and they’ve shown if you detect lameness quickly, then you’ve got a much better chance of treating it. About one in three UK dairy cows are lame, so it’s a massive problem and we should definitely be doing something about it. It’s all well trying to deal with individual cases but we need to know why these animals are going lame. Are they not comfortable in their cubicles? Are they spending too much time on their feet? Is it about how they use the space in their pen? There are a lot of different things that could cause lameness. If you monitor the animals, you can identify certain behaviors and find out if they are standing up too long or possibly something else. That’s our overall mission – to improve treatment, prevention and welfare through machine learning.”

Vet Vision AI is using deep neural networks with algorithms that function in a similar fashion to the human brain.

Dr Hyde explained: “There are millions of neurons working together to make decisions and we train these neurons to detect certain things. You show them something again and again, and they identify particular patterns. We train them on millions of labeled data points for different images that show animals and certain diseases.”

According to Dr Carslake, lifestyle tends to drive health outcomes in both humans and animals. Parameters such as nutrition, stress, housing and environment are a big factor in disease prevention. Vet Vision AI’s claims its data allows farms to anonymously benchmark their animal welfare performances against that of peers.

Although the firm’s algorithms can be trained to detect a variety of different diseases in a way that is more accurate than manual observation, Dr Hyde pointed out the technology is not designed to replace farmers but to enable them to dedicate more time to other tasks. This addresses the ongoing problem of labor shortage on farms.

Dr Carslake added: “Alongside helping vets prevent disease, Vet Vision AI’s approach is designed to help food companies gain better insights into health and welfare in their supply chains. Many companies, such as supermarkets, work with their suppliers try and measure and improve animal welfare. However, data quality is often poor as livestock health and welfare is extremely challenging to measure. In many cases, it relies on someone walking around the farm with just a clipboard which is subjective and misses important information.

“We’re developing solutions that can capture objective health and welfare insights from any farm using temporary battery powered cameras taken onto farms by vets. By processing this footage using our artificial intelligence algorithms, we’re able to collect objective and holistic health and welfare metrics from any farm. Data is then linked to a platform and integrated with current auditing databases enabling unparalleled insights into livestock health and welfare. We think this is really exciting as better measurement is key component in helping vets, farmers and the wider supply chain work together and solve problems. We believe artificial intelligence-based technology has a really important role to play.”

However, Dr Carslake believes too much on-farm data can be something of a hindrance.

He observed: “We’re drowning in data. What do you do with it? How can a farmer or a vet use this data to actually make a difference? What we’re hoping to do is tie together the computational expertise and specialist veterinary expertise in AI models that give you the data but crucially tell you what you can then do with it to make a difference by preventing disease and improving animal welfare. Farmers rebel against too much data. Just giving them more numbers won’t help, we have to drive some change.

“It flips the whole veterinary model on its head. We’re going from trying to spot a disease and treat it quickly, to knowing what the cows are doing 24 hours a day and objectively monitoring their behavior to know how we can prevent disease and improve welfare.”

Next steps

Vet Vision AI was able to establish itself as a business after receiving a grant for ￡300,000 ($384,000) from government initiative Innovate UK earlier this year. This funding helped the start-up begin building its team by hiring two computer vision engineers.

The company is looking to increase installations of its technology to support animal welfare in the cattle sector, before branching out into other food-producing animal species further down the line. The start-up has already established the ability of its technology through early-stage testing on cattle farms but will continue to do so to further validate its system. Dr Hyde said the quality of the company’s machine learning will grow and evolve, and suggested their algorithms “will only get better, the more we train them”.

The start-up has a partnership with Sainsbury’s – the second largest chain of supermarkets in the UK. This collaboration has seen Vet Vision AI’s technology adopted on 30 farms. Additionally, the company is piloting its system on around 50 other farms.

Dr Hyde – also assistant professor in computational biology at the School of Veterinary Medicine and Science at the University of Nottingham – noted Vet Vision AI has fairly immediate commercial aspirations over the next couple of years and the start-up is currently looking to source funding to fuel early uptake of its technology.

Steve Ellis has joined Vet Vision AI as chairman. Mr Ellis is the former managing director of 2 Sisters Food Group and former chief executive of the Karro Food Group. He will help guide Vet Vision AI’s commercial strategy and support its search for further investment. The start-up also retains ties with the University of Nottingham, which will help it connect with investors.

The research that formed the machine learning and computer vision fundamentals of Vet Vision AI was carried out over many years before a commercial entity was created. While the start-up’s leadership team published their research findings in academic journals and created free online tools to aid farmers, Dr Hyde noted the team also wanted to take the major step of creating a spin-out that could translate their work into a commercial business. This is sometimes a move that does not occur, with a lot of research remaining within academic settings.

Dr Hyde suggested the founders had the impetus to form a start-up due to their background as veterinary surgeons. This helped them understand the need for machine learning tools on farms and the demand there would be for a tool that could aid with disease detection and prevention.