Digital rights management of machine learning models

A system, comprising:

• one or more processors;
• and
• memory storing instructions that, when executed by at least one of the one or more processors, cause the one or more processors to perform operations comprising:
• accessing ground truth sensor data for a particular agricultural area;
• determining ground truth agricultural conditions, present in the particular agricultural area, that are correlated with the ground truth sensor data;
• determining a plurality of ground truth sensor value ranges for the ground truth agricultural conditions of the particular agricultural area indicated by the ground truth sensor data;
• generating, based on the ground truth sensor data and the plurality of ground truth sensor value ranges, synthetic sensor data that fall outside of the plurality of ground truth sensor value ranges;
• generating, based on the ground truth sensor data and the determined ground truth agricultural conditions, a plurality of synthetic agricultural conditions;
• generating a plurality of training instances including:
• first training instances each including ground truth sensor data input falling within one or more of the plurality of ground truth sensor value ranges and ground truth output corresponding to one or more of the ground truth agricultural conditions that are correlated with the respective one or more of the plurality of ground truth sensor value ranges, and
• second training instances each including synthetic sensor data input that falls outside of one or more of the ground truth sensor value ranges and synthetic output corresponding to one or more of the synthetic agricultural conditions;
• training an agricultural inference machine learning model based on the plurality of training instances to:
• generate valid agricultural inferences of agricultural conditions based on subsequently received ground truth sensor data when the subsequently received ground truth sensor data includes sensor data values that fall within the plurality of ground truth sensor value ranges;
• generate invalid or ambiguous agricultural inferences of agricultural conditions based on the subsequently received ground truth sensor data when the subsequently received ground truth sensor data includes sensor data values that fall outside the plurality of ground truth sensor value ranges
• capture, using one or more vision sensors on a farming machine in the particular agricultural area as the farming machine travels through a field of crops, images of the crops in the field in the particular agricultural area;
• apply the trained agricultural inference machine learning model to the images to generate agricultural inferences about the crops in the field;
• and
• perform, using the farming machine, one or more operational practices in the particular agricultural area, the one or more operational practices determined based on the generated agricultural inferences.

A digital rights management (DRM) method implemented using one or more processors and comprising:

• accessing, from a farming machine, an agricultural inference machine learning model that was trained previously to:
• generate invalid inferences based on sensor data that falls outside of one or more expected ground truth sensor value ranges, and
• generate valid inferences based on sensor data that falls within the one or more expected ground truth sensor value ranges;
• receiving sensor data generated by one or more sensors for a particular agricultural area for which the agricultural inference machine learning model was not trained, wherein the sensor data generated by the one or more sensors falls outside of the one or more expected ground truth sensor value ranges;
• processing the sensor data based on the agricultural inference machine learning model to generate invalid or ambiguous agricultural inferences of one or more agricultural conditions in the particular agricultural area, wherein the sensor data includes vision data captured using one or more vision sensors on the farming machine in the particular agricultural area as the farming machine travels through a field of crops;
• performing, using the farming machine, one or more operational practices in the particular agricultural area, the one or more operational practices determined based on the generated agricultural inferences.