Like the facial recognition technology now used on smartphones, a new livestock identification technology is using artificial intelligence to keep tabs on cattle.
Created by Canadian company OneCup AI, Bovine Expert Tracking and Surveillance, or BETSY, uses artificial intelligence to identify and monitor cattle through facial recognition technology.
Mokah Shmigelsky, founder of OneCup AI, grew up in the ranching community and got the idea for this technology during a family reunion. When a beef producer cousin mentioned the annoyance of losing ear tags and asked if computer vision could be used to identify cattle, Shmigelsky’s husband Geoffrey, a computer scientist specializing in artificial intelligence, asked for videos of the cattle to try it out.
Using that footage, the Shmigelskys created preliminary algorithms on the cattle in the fall of 2019. “We discovered that you could identify Black Angus just off the get-go with up to 95 per cent accuracy,” says Shmigelsky, noting that cattle with variances in hide colour and patterns are easier to identify accurately. “Since modifying our algorithms, we’re reaching 100 per cent accuracy on Black and Red Angus cattle.”
Using its 360 Live ID platform, BETSY can identify cattle “from almost any angle and at a considerable distance,” as stated on its website, using the individual animal’s appearance to create a visual ID in its system. The facial recognition ability can link to existing identification systems by reading tag numbers, and there are plans to be able to read RFID tags and link to the individual animal’s profile in the future.
Beyond animal identification, discussions with producers and industry partners helped them to determine other useful metrics to monitor, such as calving, estrus detection and health indicators. “We discovered because computer vision is a scaleable product with very little additional hardware required, just more training and algorithms, that there was more that can be done with using BETSY,” she says.
BETSY tracks different animal activities and duration of activities, including monitoring newborn calf activity. It also checks health indicators, including coughing, visual signs of depression, lameness and arching of the back. The technology also monitors nutritional intake and animal growth, and it can even be used to automatically fill out shipping manifests using an animal’s data on file.
Producers are sent four to six security cameras, based on their operation’s needs, a Wi-Fi upgrade kit and the BETSY Box, which is a tiny computer. “The cameras send their output to this box, and that’s where the brains of BETSY lie. So she either analyzes the footage right on-site with this little tiny computer, or if it needs more analyzing, then she sends that data to the cloud,” says Shmigelsky.
“We wanted to make it very easy for all producers, regardless of the technology level, to be able to use it. So all they need to do is plug it in, turn it on and she starts identifying their cows.”
Producers can access the information through their phone or computer, including live video feeds, historical video and datasets for individual animals and the entire herd. The program will send the producer text or email alerts in the event of something urgent, such as a calving issue or animals getting out.
The company has tested the technology at 20 beta sites across Canada, with some sites using this technology to monitor bison, sheep, horses and elk, and are now launching an early adopter program.
“We’re limiting it to about 100 ranches in Canada, and basically you get a discount on the system to run it for a year,” she says. “Throughout the year we’ll be adding other features that producers have identified to us as of importance, and then those will automatically get downloaded to those systems.”
Shmigelsky anticipates the full commercial launch of BETSY will take place in late 2021.
Shmigelsky is hopeful that BETSY’s monitoring abilities will be beneficial to producers throughout the year. “Because it’s so easy to implement on the ranch and there’s very little risk involved, I think it’s going to enable producers to see what’s possible in getting data and analytics on their herd and then making choices later on as to how to optimize their herd,” she says.
“Most of our producers that we’ve spoken with don’t have a lot of analytics on their animals, and this is a way where they can benefit from those analytics without having to put in the man-hours to analyze their herd themselves.”
Using biometrics to track growth
Behan’s work with infrared and multispectral cameras for livestock identification, health and growth tracking in Europe and China provided the expertise for Alpha Phenomics, a new imaging technology company.
“I said, ‘I know what we can do, we can actually look at the use of these cameras to help determine the residual feed intake, lean tissue growth rate, so that we can actually end up with more sustainable beef, converting less plant protein to more animal protein,’” says Behan, the CEO of Alpha Phenomics, who jumped at the chance to work in Canada.
Alpha Phenomics’ imaging technology allows producers to collect biometric data on livestock, including “a real-time assessment of weight, weight change over time, genetic potential of a phenotype, muscle mass index, body composition, carcass composition, indication of thermal efficiency and more,” as stated on its website.
Behan began working with this technology for a swine-related project in China, using infrared cameras “to identify when pigs were hot, and we then utilized those algorithms to dictate the ventilation rates within the buildings,” he says. This led him to work with similar cameras for swine identification, using laser point cloud grab. He describes this as “effectively grabbing a laser point image of an animal to identify it without a tag.”
Behan then explored this technology for cattle in his native Ireland, hoping to use it to accurately estimate individual animal weight. This involved the development of multispectral cameras, capturing a start-stop video of an animal moving.
“This is very important, this type of image, if you’re looking at colouration of an animal, of course, and markings and physical effects of the face or physical anomalies within the bone structure,” he says.
This also uses laser point cloud grab technology, which has one million laser points capturing 30 frames per second each when filming an animal. “Then we can actually estimate the distance between certain points on the animal linearly, so we can check the size of the animal, and then we have a very good estimation of the weight,” says Behan.
The laser point clouds build an avatar of the individual animal that provides three-dimensional, full-body biometrics to be used to track growth. “The beauty of this technology is we can measure bone — fat and lean — so we can tell you on the graph when the optimal slaughter weight is.”
In addition to research trials with the University of Saskatchewan and other groups to assess the weight of feeder cattle and increase accuracy with greater numbers, Alpha Phenomics is working on a project examining the individual growth of calves over time.
While the focus was originally growth-related, the cameras’ capabilities extend to other metrics, such as detecting disease in cattle. Infrared technology, Behan explains, can be used effectively to identify the onset of health issues.
“With cattle, we can quite easily tell you that the animal is under attack from something, and generally many days before a stockman sees clinical signs,” he says.
While the systems designed for beef are ready to go in Europe, the company is currently testing the cameras for North American use. In adapting this technology, a couple of challenges arose. One is limited internet service in parts of Western Canada not allowing for uninterrupted real-time information. Another is the cameras needing to work in extremely cold temperatures.
To remedy this, Alpha Phenomics teamed up with a Calgary-based engineering company that works in infrared measurement to rebuild the cameras to operate in these conditions. The cameras are currently being tested at academic research facilities across Canada, as well as at a farm in Ontario and in the U.S.
Behan predicts the product will be launched in January 2022. Alpha Phenomics plans to offer the technology as yearly subscription packages that include monitoring the metrics that the producer wants to measure. The technology will work with the AgSights Go360 bioTrack livestock management software, where producers can access data on their cattle in an easy-to-interpret format.
Ensuring that the technology offers value to producers in the biometric data it collects is important to Behan, who sees benefits in accurately knowing when feeder animals are ready to slaughter or when the first signs of disease in cattle appear.
“We were originally working with animal identification, identification of weight, growth curves, but we think for the beef ranchers we need really simple technologies that he will use and will get a value from.”