Russia’s Peshchanokopsky agriculture group has allocated part of its croplands in the south of the country for a very special training ground. It will be used to train AI-based agricultural machinery developed by Cognitive Pilot.
The company is also working on the world’s first Open Dataset – an image database that will be used to train autonomous agricultural machinery – a tool that will facilitate the development of AI-based autonomous driving systems for agricultural machinery, Cognitive Pilot CEO Olga Uskova commented for Invest Foresight.
“The training facility enable the simulation of a wide range of situations, including those extremely rare in real agricultural work, the different weather conditions, different lighting – something that has enormous value for building the database and ultimately ensuring the safe operation of an autonomous agricultural machinery control system,” she said.
Vladimir Yalovenko, the main shareholder and member of the Board of Directors of the agribusiness, said:
“In turn, we will add our expertise and subtleties of agricultural business processes to the project.”
It should be underscored that until recently, developers never focused on building global data sets for agriculture. At best, they limited their efforts to addressing local problems like, for example, spraying weeds. In this sense, the Cognitive Pilot project is truly the world’s first attempt to create a universal database that can then be used around the world.
What will the dataset include? A rather big scope of data. First and foremost, it will contain footage from cameras installed on tractors and combined harvesters. The agricultural equipment will also be fitted with mileage and trajectory meters. Other data collected will be weather conditions, geometrical characteristics of farming land and cultivated crops.
‘Field’ datasets are the most representative as they are collected literally in the field rather than through computer modeling. A more accurate picture allows taking into account the maximum number of parameters required for the system’s steady operation.
It is noteworthy that Russia is only one of the company’s several platforms for testing this technology. Currently, it is expected that this year it will expand to such countries as Canada, the United States, Argentina, South Korea and China. Although developers themselves admit that the coronavirus pandemic may make adjustments to their plans but they hope it will not be a serious obstacle.
“Our goal is to create the largest and most representative database for smart farming in the world,” Olga Uskova summarized in a comment for Invest Foresight.
There is only one problem. Who do you ask to plow your garden in spring now? Before, you could ask the local tractor driver you know. Now, I don’t think you will be able to get AI to do it for you.
By Taras Fomchenkov