Difficulty sample acquisition method, device and equipment and readable storage medium

Abstract

The embodiment of the invention provides a difficult sample acquisition method, device and equipment and a readable storage medium, and after second electronic equipment identifies a target image to obtain a second identification result, the target image and the second identification result are sent to first electronic equipment. And the first electronic equipment identifies the target image to obtain a second identification result, and determines whether the target image is a difficult sample according to the first identification result and the second identification result. By adopting the scheme, the same image is identified through different electronic devices to mine the difficult sample, the coverage range of the difficult sample is expanded, and the difficulty and workload of mining the difficult sample are reduced.

Description

technical field [0001] The present application relates to the technical field of deep learning, and in particular to a method, device, equipment and readable storage medium for acquiring difficult samples. Background technique [0002] Currently, deep learning is an important branch of machine learning technology. Deep learning is the process of continuously training the training data set to obtain a deep learning model. The quality of a deep learning model heavily depends on the number of training samples in the training dataset and the quality of the training samples. [0003] Generally speaking, training samples include easy samples and hard samples. When the number of simple samples in the training data set reaches a certain level, the accuracy of the deep learning model mainly depends on the difficult samples. In order to obtain difficult samples, a common way is to adjust the weight of samples in the training data set to find difficult samples. [0004] The above m...

AI Technical Summary

Problems solved by technology

[0004] The above method of finding difficult samples is based on a fixed training data set, and the coverage of a fixed training data set is often limited

Therefore, the above-mentioned difficult sample discovery method cannot guarantee the coverage of difficult samples, which leads to low accuracy and poor scene adaptability of deep learning models.

Images