Method and device for determining training sample and method for training deep learning model

Abstract

The invention provides a method and device for determining a training sample and a method for training a deep learning model. The method for determining the training sample comprises the steps: acquiring the feature vectors of N to-be-labeled samples in a first to-be-labeled sample set, wherein N is a positive integer; and determining m to-be-labeled samples from the N to-be-labeled samples according to the difference between the feature vectors of the N to-be-labeled samples, so as to label the M to-be-labeled samples and obtain a labeled sample set, wherein M is a positive integer and is less than N. The to-be-labeled samples are selected according to the difference between the feature vectors of the N to-be-labeled samples, so that the samples can be screened from the to-be-labeled sample set more efficiently to be labeled, the number of training samples needing to be labeled is reduced, labeling resources are utilized more effectively, and the performance of the deep learning modelis improved.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a method and device for determining training samples, and a method for training a deep learning model. Background technique [0002] In recent years, deep learning techniques have transformed computer vision and have found application in a plethora of consumer-facing products. For example, in the field of medical imaging, the segmentation of medical images requires high precision. Thanks to the development of deep learning technology, it has achieved excellent results beyond traditional segmentation methods, which is of great significance for clinical analysis, diagnosis, treatment and prognosis. . [0003] However, an effective deep learning model usually requires a large number of high-quality labeled training samples, and the labeling workload is heavy. Moreover, the labeling of training samples is a very time-consuming and labor-intensive task. For example, the segmen...

AI Technical Summary

Problems solved by technology

[0003] However, an effective deep learning model usually requires a large number of high-quality labeled training samples, and the labeling workload is heavy.

Moreover, the labeling of training samples is a very time-consuming and labor-intensive task. For example, the segmentation labeling task needs to manually outline the edge of the target, and the labeling of medical images requires corresponding clinical knowledge to label accurately, and the labeling cost is high.

Images