Human brain waves are captured by AI to identify diseased plants

Some scholars from the Brazilian Agricultural Research Corporation – Embrapa, in partnership with two other technology, created a way, through artificial intelligence, to analyze changes in plants with excellence.

In the past, this would only be possible if specialists performed a manual exam. Check out more information about the AI ​​that identifies diseased plants below.

Technological advances in agriculture

The technology makes it possible to automate labeling with annotations of classification data and through the identification of diseased plants through images. With this new technology, the researchers hope to speed up decisions taken in the field, in order to reduce production losses and bring about great improvements in the use of natural resources.

The first results of the experiment

On Tuesday (07/02) the results of the first test of the experiment, carried out with soybean leaves, were released. According to the researchers, this technology could be applied to other types of productions.

This diseased plant identification system can also be added to agricultural machinery, drones and cell phones.

All the preparation for this work began in 2022 and had the help of the companies Macnica DHW, which is part of the South American operation of the Japanese group Macnica Inc., and InnerEye, from Israel.

For the tests of the experiment, two diseases that cause symptoms on soybean leaves were chosen, such as Asian rust and powdery mildew, prominent in southern Brazil. With the use of the equipment, they were able to differentiate diseased leaves from healthy ones.

Asian soybean rust is the most serious of the crop's diseases, and can lead to up to 80% losses if not controlled. The costs of caring for this disease exceed US$ 2 billion per crop in Brazil, according to a survey by the Antiferrugem Consortium.

How the experiment works:

• The neural signals of two specialists from Embrapa Soja were captured by a helmet with electrodes;
• Both evaluated about 1,500 images of diseased and healthy leaves for the tests;
• Through the capture of brain waves, the experiences, judgments and classifications they make of the images observed by the specialists were identified;
• The models generated from the spontaneous electrical currents emitted in the experts' brains and the software allowed training the algorithm of standard artificial intelligence to bring it ever closer to the criteria used by specialists in the visual identification of diseases in plants;
• After capturing the brain signals that specialists emitted when viewing images of diseased plants, the software automated the diagnosis, making the step faster and more efficient;
• The results obtained comparing the post-training algorithms indicated that they are practically at the same level of precision as the experts.