Button Text
Dec 6, 2019
Min Read

AI in Agriculture: Changes for the Better

Share on TwitterShare on Twitter
Share on TwitterShare on Twitter
Share on TwitterShare on Twitter
Share on TwitterShare on Twitter
Christopher Marshall
Data science technical writer

Agriculture: keeping civilizations sedentary for millennia. Throughout it all, technological innovation has guided the practice of cultivating plants and livestock towards higher and higher yields. Now, this storied and globally essential industry is on the brink of a new era. Artificial intelligence—more specifically machine learning (ML)—offers enormous untapped potential to further improve yields while reducing resource use and limiting negative environmental impacts. It’s the dawn of a more efficient and more humane way to farm. From bovine mass surveillance to precision weed elimination, if you’re involved in agriculture, you best be prepared to get digital.

Food demand is forecast to increase 59–98% by 2050. This kind of increase requires either more land or innovative solutions to increase both efficiency and output. How can ML help? In this overview, we’ll focus on three areas of agriculture:

  • Crops
  • Livestock
  • Water and soil

Let’s explore how ML is already being used in each and what the future might hold.


The most significant way ML is impacting crop management is through yield estimation and evaluation. Computer vision techniques can be applied to crops to evaluate which are ready to harvest, allowing farmers to better plan their activities. ML can also be applied more generally, taking into account historical data to determine the best time to harvest and approach buyers in order to better balance the market and maximize profits. Computer vision can also reduce waste and improve the value of a harvest by categorizing the produce and examining it for flaws and contaminants.

Fertilizer use can be made more resourceful: at super.ai, we’ve worked with Triputra Group, an Indonesian agriculture conglomerate, to evaluate their crop using drone footage. They were looking to economize their fertilizer use. Our solution was to implement a computer vision tree count and canopy size estimation. In addition, we help them automate bunch counting: workers harvesting the trees cut off branches and photograph them. We used image analysis techniques to estimate the yield and payment for the workers.

Weeds form a significant threat to crops, but identifying and combating them is a difficult, time-consuming task. ML can locate weeds hiding among crops and determine their species, allowing for the deployment of specialised tools that forgo the need for indiscriminate herbicide use. 

Likewise, diseases and pests affecting crops are traditionally combated through the wide dissemination of pesticides. This is a costly as well as environmentally detrimental practice. ML allows for a new approach: real-time targeted identification and eradication of pests and diseases through highly localised pesticide use and early detection of diseases in the crop lifecycle, improving yield and quality.


We’re all familiar with facial recognition, with many of us using it multiple times a day to unlock our phones. Now, we’re beginning to turn this technology on a new face, that of the cow. Bovine facial recognition is being used to identify and track individuals within a herd without the need for radio frequency identification (RFID) tags. This technology is of course expanding beyond the bovine, in the direction of the pig, the sheep, and beyond. The ability to track and analyze behavior patterns—how much time the animals spend standing, lying, moving, etc.—on the level of individual animals can help determine the health of the individual, as well as the overall health of the herd. ML can also spot warning signs of stress and disease, leading to improvements in welfare and productivity. 

Herd productivity can be further improved upon through predicting future yields in, for example, eggs, milk, and animal weight. This information can help better optimize the diet on which the animals are reared as well as the environment within which they’re kept.

Water and soil

Understanding precipitation and evapotranspiration (the process of water transferring from the land and plants into the atmosphere) is essential to resourceful agricultural management. The processes involved are complex, with temperature, relative humidity, solar radiation, and wind speed all playing a role. 

Predictive analytics can provide estimations of precipitation and evapotranspiration. Pair this with soil samples and other data and you can train ML models to provide powerful insights into soil moisture levels, temperature, and overall condition. Farmers can harness this information to irrigate their crops more efficiently, to the benefit of both profit margins and the environment. Such systems can continually monitor farm conditions automatically, leading to a less work-intensive, more resourceful form of agriculture.

How to get started with ML in agriculture

The wealth and complexity of the data that agriculture offers makes it a fascinating industry for ML. Not only that, but agriculture’s expected increase in demand means that the time is ripe to harvest that data and turn it towards transforming agriculture into a more efficient, less labor-intensive, and more profitable industry. 

If you work in agriculture and you think that your business could benefit from ML, or you want to know what ML can do you for you, reach out to us.

We will continue exploring specific agricultural use cases in detail in future posts, so stay tuned.

Other Tags:
Share on TwitterShare on Twitter
Share on FacebookShare on Facebook
Share on GithubShare on Github
Share on LinkedinShare on Linkedin

Get a customized demo with your documents

Book a free consultation with our experts.

You might also like