Microscope system and method for monitoring learning process of machine learning model

Abstract

A microscope system and method for monitoring a learning process of a machine learning model are described. The microscopy system comprises a microscope (80) having a camera (82) for capturing a microscopic image (4) and a computing device (85). The computing device (85) processes the microscope image (4) by means of a machine learning model (M). The learning process of the machine learning model (M) is performed by the training system (90). During the learning process, model parameter values (P1-P9) of the machine learning model (M) are adjusted using the training data (T). During a learning process, a quality metric (Q1) based on training data (T) and a quality metric (Q2) based on verification data (V) are calculated for current model parameter values (P1-P9), respectively. The training learning progress (1) and the verification learning progress (2) are formed from quality metrics (Q1, Q2). The training system (90) comprises a verification model (10) which is input in a learning process a training learning progress (1) and a verification learning progress (2) in the learning process. The verification model (10) is designed to generate a quality assessment (17) of the learning process as a function of the training learning progress (1) and the verification learning progress (2).

Description

technical field [0001] The present disclosure relates to a microscope system using machine learning models. The model parameter values ​​for a machine learning model are defined during the learning process. The present disclosure also relates to a method for monitoring the learning process (training progress) of a machine learning model, in particular for microscopes or other measuring or image processing devices. Background technique [0002] Machine learning models and especially deep learning methods are playing an increasingly important role in product development, especially in the development of microscopes. For example, modern microscopes use machine learning models to automatically process overview images for sample navigation. Regions of interest, sample containers / vessels and sample carriers can be identified in the overview image by segmentation, classification and detection methods. This enables automatic further processing or precise navigation, eg by sample ...

AI Technical Summary

Problems solved by technology

On the other hand, the generator is trained to produce an output where the discriminator is ideally indistinguishable from the training data

Images