Target detection method and device

Abstract

The invention discloses a target detection method and device, and the method comprises the steps: setting at least one adjustment mode for a downsampling structure of a YOLO-v4 backbone network based on the features of a to-be-detected target; adjusting a down-sampling structure of the YOLO-v4 backbone network by using an adjustment mode, and constructing a target detection model based on YOLO-v4; inputting the detection image into a target detection model, extracting a down-sampling feature map of the detection image by the target detection model, and obtaining a target detection result according to the down-sampling feature map; determining the size of the down-sampling feature map according to the adjusted down-sampling structure. The technical scheme has the beneficial effects that the precision of target detection can be improved by adjusting the down-sampling structure, images of scratches, wool fibers and the like are linear and small-size targets by taking an industrial defect detection scene as an example, and if an original down-sampling structure is used for processing a detection image, the detection performance can be obviously reduced by multiple times of down-sampling. The improved target detection model effectively solves the problem.

Description

technical field [0001] The present application relates to the technical field of computer vision, and in particular to a target detection method and device. Background technique [0002] YOLO (English full name You Only Look Once, no Chinese name in the industry) is a typical single-stage target detection technology, that is, it directly returns information such as the position and category of the target based on the original image. It has now developed to the fourth version, namely YOLO-v4. figure 1 A schematic diagram of the network structure of YOLO-v4 is shown. It can be seen that it contains a downsampling structure composed of multiple downsampling layers, but this setting has some shortcomings. For example, in industrial defect detection scenarios, some defects are still It is difficult to accurately identify, and there is still room for improvement in this technology. [0003] It should be noted that the statements herein only provide background information related...

AI Technical Summary

Problems solved by technology

figure 1 A schematic diagram of the network structure of YOLO-v4 is shown. It can be seen that it contains a downsampling structure composed of multiple downsampling layers, but this setting has some shortcomings. For example, in industrial defect detection scenarios, some defects are still Difficult to identify accurately, there is still room for improvement in this technology

Images