An app that identifies weeds, insect damage and diseases could change the way crop scouting works.
“The reality in field crop scouting for pests today is that looking for disease, weeds and insects is done often with textbooks and apps that make use of manual menus and you compare pictures,” said Warren Bills, business development manager in Canada for xarvio, an agriculture technology company backed by Bayer CropScience. “There must be a better way and a more efficient way to identify pests, quickly and more accurate.”
The app launched recently in Canada and is free to download to smartphones.
The company has been working for two-and-a-half years on the algorithm, and had to give the system time to make sure it would recognize enough crop pests to be useful.
As thousands of images of crops and weeds are sent to the database, it learns and will have greater accuracy over time.
The app was launched in other markets in late 2017 and now has close to 60,000 users around the world.
The global reach means that year-round, weeds are being uploaded to the database, now housing more than 100,000 images.
At this point the app is more capable of the most common weeds, but as time goes on the accuracy will improve. At this point it will give a percentage accuracy for the evaluation.
“We believe it will really scale globally. Any picture taken in the U.S. or Canada or China will help the algorithm improve,” said Bills. Once there are 500 to 1,000 images of a particular weed in the database, the accuracy is strong.
In order to build the database, xarvio has connected with some Canadian universities, including the University of Alberta, with samples of problem weeds. They have added images of weeds to the database.
Users find they have an “aha” moment, said Bills, when they are able to identify a problem in a crop and that’s when trust is built in the system.
Users are able to create scouting trips where pictures and notes created can be stored together.
The app doesn’t yet tie the information into larger crop-management systems. Bills said the goal was to keep the app response and simple, although he said that integration, or the ability to export the data, could happen in the future if it makes sense.
The app also shows regional trends in crop diseases and insect pressure, not specific to individual farms, but regionally, so a user can tell what pressures are growing or present in their area.
Bills said the longer-term goal is to be able to instantaneously identify weeds, with the potential to put the system on a sprayer so that it would only spray for certain pests as needed.
“This is not about replacing the great knowledge base out there, it is about extending it,” he said.