Data enhancement method, device, computing device and computer storage medium

Abstract

The invention discloses a data enhancement method, device, computing equipment and computer storage medium. Wherein, the method includes: randomly generating N sets of mask data, generating a random mask based on the N sets of mask data; performing background extraction processing on the training image to be enhanced to obtain the corresponding first background image and first foreground image; Background extraction processing to obtain the corresponding second background image; according to the random mask, perform background splicing processing on the first background image and the second background image to obtain a third background image; superimpose the first foreground image and the third background image Processing to obtain data-enhanced images. By diversifying, enriching, and complicating background images, the target detection model trained by using data-enhanced images can also have stronger adaptability under the condition of limited training image samples. Effectively improve the generalization ability of the target detection model, and greatly improve the accuracy of the target detection model.

Description

technical field [0001] The present invention relates to the technical field of data processing, in particular to a data enhancement method, device, computing equipment and computer storage medium. Background technique [0002] In recent years, object detection models trained based on deep learning have achieved great success in image object detection tasks. The performance of object detection models relies heavily on a large number of training image samples, however, in real-world applications, it is difficult to obtain a large number of effective training image samples. [0003] In the image target detection of deep learning, there will be the following situation: in the case of a small training data set, the trained target detection model has a high accuracy rate on the training set, but the accuracy rate in the test set or real application Relatively low, the generalization ability of the object detection model is poor. In order to overcome the above-mentioned problems,...

AI Technical Summary

Problems solved by technology

Most of the common data enhancement methods are by performing perspective transformation, affine transformation, adding noise, adjusting color brightness, etc. on the entire image, or mixing multiple images with different weights. These data enhancement methods It is difficult to produce the expected effective effect in the image target detection task

Images