Can you ever imagine that the future of the earth’s food crops isn’t tied to fertilizers or organic technology, but, on AI or artificial intelligence? Food security is constantly threatened by a never ending list of threats. Climate variability in weather-unstable regions can cause sudden droughts that can kill off potential harvests. Also, political turmoil is also cause for logistical blockades that can block off farming, harvesting, and shipping produce.
Then there is also the sudden appearance of plant disease and insect invasions can wipe out entire crops. A team of researchers from the École Polytechnique Fédérale de Lausanne in Switzerland and Pennsylvania State University in the U.S. have turned towards artificial intelligence to aid in agriculture.
This means deep learning algorithms have now been trained to detect crop disease before it can spread. In all developed countries, large scale crops are farmed through large-scale operations, backed up by sufficient finances, manpower, and other materials needed to help fight crop disease early on.
But in developing and third world countries, up to 80 percent of agricultural production is conducted by smallholder farmers. These small-scale operations are more prone to the devastating effects of crop disease that can wipe out crops in a matter of weeks and can lead to widespread famine.
Unfortunately, 50 percent of the world’s poor population lives in smallholder farm households that have little or no resources to address crop diseases. The new Artificial Intelligence is called Machine Vision and has already been proven to excel in driving cars autonomously, diagnosing cancer, and pinpointing specific persons in photographs.
It is now “ripe” to be developed in aid for agriculture. According to David Hughes, co-author of the research and Penn State professor, along with lead author Sharada Mohanty and co-author Marcel Salathe of EPFL, if Machine Vision can be trained as a game changer to drive cars, go after cancer cells, and do faces, then it can certainly do plant diseases.
Video Courtesy of YouTube:
The researchers developed a program that’s fast, efficient, and compact enough to pack into a smartphone. They then trained the algorithm by feeding it huge datasets gathered from Plant Village, an open access online archive of plant photos including images of plant disease.
With this data the researchers trained the algorithm to identify 26 different diseases in 14 different plant species using over 50,000 photographs from the online archive. According to the researchers, after the training phase, the algorithm performed with 99.35 percent accuracy, giving smartphone users the ability to identify diseases like an expert, something that the ordinary eye may not spot immediately, if at all. This is another reason, that Artificial Intelligence will literally transform our society in the coming years.
But the researchers are not stopping to massage their laurels as they continue to constantly refine and add more data into the algorithms. It is hoped that in the coming months the new program can be launched in an app on any smartphone. The world is now racing towards nine billion people and feeding them all is becoming a daily challenge, but with scientists doing work like this perhaps their crucial effort will not go unnoticed.