The AI platform can be used to both count the number of livestock or monitor the health of the animals as needed.
Livestock farming is big business, and now Plainsight has introduced an AI platform for assisting farmers with ensuring their livestock remain present in their correct numbers. The platform, known as Vision AI, has the capability to pinpoint the number of any particular livestock, whether it be cattle, sheep or pigs in a given area, with 99.7% accuracy, according to Plainsight.
SEE: Artificial Intelligence Ethics Policy (TechRepublic Premium)
“We had customers come to us with this problem, and that’s typically how we build solutions,” said Elizabeth Spears, Chief Product Officer at Plainsight. “When you have situations where livestock aren’t being counted accurately, you can be susceptible to fraud in the transportation within that process. There’s a number of places where, along the supply chain, the counts of livestock can become inaccurate.”
Vision AI for livestock management
With cattle alone ranging between $800 and $1,200 per head, it is imperative that those in the farming business know exactly how many of a particular animal they have on their property. While a large industrial farm was the inspiration for the artificial intelligence platform, Spears says that this feature can be easily integrated into both large scale and smaller family farms as well.
“I would say it could work for both, but it’s most applicable for the larger farms, because usually that’s where counting becomes the most economically essential,” Spears said. “It’s a very practical solution. It’s also very affordable. It’s the most affordable for larger scale operations, and the way that it works is we typically work with the company to get it installed. So with Vision in particular, depending on your environment, you need to maintain and update the model.”
To help combat this problem, Plainsight’s object counting system runs on a number of different components:
- An accurate object detection model to classify and localize where the object is inside each frame of a video. This usually requires training a custom model on a labeled dataset.
- An object tracking algorithm to track where each individual object moves to from frame-to-frame.
- A registration zone where object detection and tracking is applied when objects enter the camera view.
- A counting line that triggers the object count as each crosses it.
- An object that moves backward can be counted multiple times if you do not account for movement direction.
- A deregistration zone where object detection and tracking can be confidently removed after the counting has happened.
Besides simply detecting the correct number of a type of livestock, the platform also aids farmers in ensuring that animals remain healthy and secure within the farm.
“Counting is one thing that it can do, but it can also really help with kind of monitoring the health of the animal and also with sort of overall operations on the producer’s side,” Spears said. “You can look at animal gait, so how they’re walking as an indicator of animal health, or food intake, and there’s other kinds of visual indicators of health. It creates a way for producers to be able to manage the larger numbers of livestock in general.”
Spears also says that if industrial or smaller scale farms would like to install the Vision AI platform for their livestock management needs, the learning curve would be minimal.
“The platform shows what’s going on with your data really clearly, and that’s really essential for being able to monitor these over time,” she said. “One of the reasons that people can have trouble with the long term accuracy of their models is not being able to have really clear understanding of the data that’s coming in and the detections that are coming in. Accuracy is really essential, especially for these larger scale businesses, because miscounting the cattle is a tens of millions of dollars problem for them a year.”