DETECTION OF PLANT DISEASES USING MULTI-STEP, MULTI-SCALE DEEP LEARNING Russian patent published in 2024 - IPC G06F18/2431 G06V10/82 

FIELD: physics.

SUBSTANCE: group of inventions relates to image recognition and can be used to recognize plant diseases. Recognition method comprises the following steps: obtaining a new image from a user device by a processor; application by a processor of a first digital model trained to assign a plurality of first regions within a new image to a plurality of classifications, which are selected from a first group of plant diseases having symptoms with a size within a first range, a healthy state or a composite class for a combination of plant diseases having symptoms of other sizes; application by a processor of a second digital model trained to assign one or more second regions, which are assigned by the first digital model to a prefabricated class, to one or more classifications, which are selected from a plant disease, having symptoms within a second range, a healthy state or a composite class for a combination of a plurality of plant diseases; transmission of classification data to a user device.

EFFECT: high accuracy of recognition.

15 cl, 12 dwg