Descartes Labs, a startup with image recognition technology that can help companies answer questions related to agriculture, is announcing today a $5 million round of funding.
The startup can work with satellite images to predict what will happen next. To do that, Descartes Labs depends on a type of artificial intelligence called deep learning. The method entails training systems called artificial neural networks on lots of data, like a whole bunch of satellite images containing different things, and then getting the neural networks to make an inference on the contents of new images.
Using this technology, Descartes Labs claims that it can make more accurate predictions than the U.S. Department of Agriculture.
The USDA estimates that the corn harvest will be 13.6 billion bushels this year, which works out to a bit over a million billion corn kernels. Each kernel weighs about 280 milligrams. Descartes Labs is able to predict the corn harvest to an accuracy near 99%, which means we are using satellites orbiting above the corn fields to effectively measure the weight of the average kernel of corn to better than 3 milligrams, an error which is roughly the weight of a grain of sand.
This isn’t the first deep learning startup to take a vertical-specific approach. Enlitic focuses on health care. Meanwhile, the more general-purpose Ersatz Labs recently decided to suspend development of its cloud service and application programming interface (API) and is now focusing on consulting.
Descartes Labs spun out of the Los Alamos National Laboratory in New Mexico last year and is proudly based in Los Alamos, New Mexico, with around 13 employees.
Cultivian Sandbox led the new round in Descartes Labs. Crosslink Capital, Data Collective, TenOneTen Ventures, and ValueStream Labs also participated.
To date the startup has raised $8.78 million. A spokesperson declined to talk about how many paying customers it has.