Sample image processing method and device, equipment, and storage medium

Abstract

The embodiment of the invention discloses a sample image processing method and device, equipment and a storage medium. The method comprises the following steps: firstly, obtaining a plurality of first target frames of a sample image, the first target frames comprising a first target image, the first target image comprising channel features and spatial features; calculating the first target images included in each first target frame by using a channel feature calculation module and a spatial feature calculation module included in a preset neural network to obtain a feature-enhanced target image set; and finally, adding the target image set subjected to feature enhancement into the sample image to obtain an enhanced image. According to the embodiment of the invention, the invention solves a problem that a subsequent detection model cannot accurately detect the target image because of the poorer lifting effect of the target image in a conventional sample image processing scheme, improves the features of the target image included in the sample image, and enables the subsequent detection model to accurately detect the target image.

Description

technical field [0001] The present invention relates to the field of image processing, in particular to a sample image processing method, device, equipment and storage medium. Background technique [0002] At present, for the safety of self-driving vehicles, in order to obtain more comprehensive feature information in high-definition images, it is very important to detect target images from high-definition images. [0003] In order to detect target images in high-definition images, image enhancement methods are usually used at present. [0004] However, in the current image enhancement methods, by using difficult negative samples as compensation to supplement positive samples, only using the original data set, there is no obvious improvement effect on small target images. [0005] In addition, by fusing features of different scales, while expanding the receptive field, features of multiple scales are fused to enhance small target images. Because the problem is not consider...

AI Technical Summary

Problems solved by technology

[0007] The embodiment of the present invention provides a sample image processing method, device, equipment and storage medium, which solves the problem that the subsequent detection model cannot be accurately detected due to the poor effect of target image enhancement in the existing sample image processing scheme. The problem of the target image is enhanced, and the characteristics of the target image included in the sample image are enhanced, so that the subsequent detection model can accurately detect the target image

Images