Industrial part detection method based on YOLO v8 algorithm

Abstract

The invention discloses an industrial part detection method based on a YOLO v8 algorithm, and the method comprises the following steps: S1, obtaining an original image of an industrial part, and carrying out the preprocessing of the original image, and obtaining a data set; s2, a YOLO v8 model is improved; s3, sending the industrial part picture set into the improved YOLO v8 network model for training, outputting a detection result, and evaluating the detection result; the improvement of the YOLO v8 model comprises the following steps: S21, using an SPPELAN module to improve an SPPF; s22, a C2f part of the YOLO v8 is improved by using a GhostModule provided by the CVPR2024-DynamicConv, and the C2f part of the YOLO v8 is improved by using a C2f part of the YOLO v8; and S23, an improved loss function Quity FocalLoss is used, and the Quity FocalLoss supports and dynamically adjusts the difficulty through a joint representation method and continuous labels, so that the model pays attention to samples which are difficult to position or classify. The invention provides an industrial part detection method based on a YOLO v8 algorithm, and aims to solve the technical problems of low defect detection precision and low detection efficiency of industrial parts in the prior art.

Images