SYSTEM AND METHOD FOR AUTOMATIC MACHINE LEARNING (AutoML) OF COMPUTER VISION MODELS FOR ANALYSING BIOMEDICAL IMAGES Russian patent published in 2023 - IPC G16H30/00 G06N20/00 G06T7/00 

FIELD: computing technology.

SUBSTANCE: invention relates to a system and a method for automatic machine learning (AutoML) of computer vision models for analysing biomedical images. In the method, the biomedical image data required for testing, training and validating computer vision models for the biomedical image analysis are automatically downloaded; the biomedical image data are converted into a format accepted for searching, training, and assessing; AutoML search for computer vision model architectures for the biomedical image analysis is conducted using the training sample and the test sample based on the biomedical image data; the computer vision models for the biomedical image analysis with the architectures found are trained using the training sample, wherein the best of said trained models is selected and transferred for assessment, the criterion for selecting the best model herein being the model achieving the set values of one or more model metrics when tested using the test sample; the best computer vision model for the biomedical image analysis selected is assessed using the validation sample based on the biomedical image data, wherein the validation sample is based on the data not contained in the test sample or the training sample.

EFFECT: higher accuracy of analysing biomedical images due to the most effective computer vision model being determined.

10 cl, 6 dwg