Method and device for generating training image set in target detection and semantic segmentation task

Abstract

The invention provides a method and device for generating a training image set in a target detection and semantic segmentation task. The method comprises the steps: carrying out the segmentation processing of a to-be-detected large image to obtain segmented images which comprises segmented images containing target features and background sample images not containing the target features; taking the segmented images containing the target features as defective samples and adding the defective samples to a training set; performing region division on a to-be-detected large image to obtain a background image region; in the background image region, randomly selecting background sample images, adding the selected images to the training set, wherein the number of the background sample images is calculated according to the number of defective samples and the background proportion of the to-be-detected large image. According to the method, the segmented images containing the target features and the randomly selected background sample images are used as the training image set for training, so that the problem of low training efficiency caused by a large number of training images when all the background sample images are input into the training image set is avoided.

Description

technical field [0001] The present application relates to the technical field of image target detection, and in particular to a method and device for generating a training image set in target detection and semantic segmentation tasks. Background technique [0002] In the field of image processing, target detection algorithms (Detection) and semantic segmentation algorithms (Segmentation) are generally used to extract target features of labeled images. Target detection algorithm (Detection) and semantic segmentation algorithm (Segmentation) are two very important directions in the field of deep learning computer vision (CV). The target detection algorithm extracts target features by training a series of labeled images and uses them to Predict whether the newly input image with the label contains the target feature; if it contains the target feature, the target detection algorithm will mark the target feature through a rectangular detection frame. The semantic segmentation al...

AI Technical Summary

Problems solved by technology

[0005] This application provides a method and device for generating a training image set in target detection and semantic segmentation tasks, to solve the problem in the prior art of inputting all the segmented small images into the target detection and semantic segmentation model for training, which will The training cycle is too long, which affects the training efficiency of target detection and semantic segmentation models

Images