Precision agriculture tools developed through open source frameworks are enabling researchers17 and citizen scientists18 alike to build upon, adapt, and customize free software for their own specific needs. From Tanzania21 and Japan22 to the U.S., the agricultural community is leveraging data and benefiting from free machine learning platforms, like TensorFlow and CaffeNet, to improve the health of their plants and animals while maximizing productivity.
Whether it’s farmers relying on Farm Hack to share DIY tech tools or home gardeners using FarmBots to plant seeds and kill weeds in their backyards, these collaborative innovations are changing the landscape of food production in the digital age. And while many small and medium-sized American farms have yet to fully adopt machine learning tools, that may be changing. Dairy farmers in the U.S. have been facing a crisis, with milk prices at an all-time low.23 Hart Dairy Farm in Waynesboro, Georgia is the first U.S. farm to develop and implement a machine learning application using TensorFlow for tracking their livestock. Their application, called IDA, continues to improve over time by regularly collecting and analyzing data from sensors attached to cow udders. By using remote-control sized transmitters, owners can know when their Holsteins or Guernseys are chewing cud, feeling sick, or ready for insemination. The volume of data captured and processed by Hart’s software is astounding: originally trained on the equivalent of 600 years of cow data, it can replicate that volume every two-and-a-half months or so, given that the equivalent of eight years of cow data is created and collected every day.24
Melissa Brandao, CEO and founder of HerdDogg, originally developed smart herd monitoring tools to be a kind of “Fitbit for cows,” measuring everything from a cow’s ear temperature to their activity levels via a Bluetooth-enabled tag clipped to their floppy ear. (Brandao says she focused on developing ear tags because it’s the only “real estate” on a cow that developers have been granted access to.) Using this data, the tag can then tell farmers or researchers whether cows are in heat, missing, or about to become sick.
In addition to the ear clips, HerdDogg manufactures what they call DoggBones— small, passive readers that can collect data about a herd anytime and anywhere, no WiFi or cellular connection required. Combining both allows farmers to get a granular look at the lives of their dairy-makers, giving the average small farmer access to as much as 100,000,000 rows of data per year.