MSFT-YOLO: Improved YOLOv5 Based on Transformer for Detecting Defects of Steel Surface

Abstract

With the development of artificial intelligence technology and the popularity of intelligent production projects, intelligent inspection systems have gradually become a hot topic in the industrial field. As a fundamental problem in the field of computer vision, how to achieve object detection in the industry while taking into account the accuracy and real-time detection is an important challenge in the development of intelligent detection systems. The detection of defects on steel surfaces is an important application of object detection in the industry. Correct and fast detection of surface defects can greatly improve productivity and product quality. To this end, this paper introduces the MSFT-YOLO model, which is improved based on the one-stage detector. The MSFT-YOLO model is proposed for the industrial scenario in which the image background interference is great, the defect category is easily confused, the defect scale changes a great deal, and the detection results of small defects are poor. By adding the TRANS module, which is designed based on Transformer, to the backbone and detection headers, the features can be combined with global information. The fusion of features at different scales by combining multi-scale feature fusion structures enhances the dynamic adjustment of the detector to objects at different scales. To further improve the performance of MSFT-YOLO, we also introduce plenty of effective strategies, such as data augmentation and multi-step training methods. The test results on the NEU-DET dataset show that MSPF-YOLO can achieve real-time detection, and the average detection accuracy of MSFT-YOLO is 75.2, improving about 7% compared to the baseline model (YOLOv5) and 18% compared to Faster R-CNN, which is advantageous and inspiring.

An Analysis of Defect Detection on Steel Strip Surface Using Modified Yolo V5

“…Researchers started concentrating on the topic of deep neural networks and how to apply them to object detection after Krizhevsky developed AlexNet in 2012 to obtain outstanding results in image classification [38] [36]. Machine vision technology in convergence with artificial intelligence has been rapidly improving and is helping agricultural, industrial, medical and other complex real-time applications [3][4] [5] [6].…”

“…Convolution networks lacks the capacity to gather global contextual information. In large neighbourhood background and to improve semantic discriminability, In order to give higher levels of detection features for detection by merging global information, the TRANSFORMER module is introduced [36]. TRANSFORMER module delivers concrete results in image classification and object detection.…”

“…Steel surface defect images at times affected by production environment which stipulates cluttered background of defect images. Using this TRANSFORMER module richer defect contextual features are provided that gathers various positioning information for the entire model [36]. The Convl in the last block of backbone module of YOLOv5 is replaced by C3TR.…”

NRT-YOLO: Improved YOLOv5 Based on Nested Residual Transformer for Tiny Remote Sensing Object Detection

“…Thus, we attempt to address the tiny object detection problems in remote sensing imagery by utilizing YOLO, the onestage detector. YOLOv5 has shown its huge potential in detecting tiny objects [23][24][25][26]. By introducing Bi-FPN, [26] it enhances the feature extraction ability of the network.…”

“…For nightmare remote sensing imagery detection, [25] improves the accuracy using a series of special data augmentation. References [23,24] utilize Transformer components to combine global information with image features, and satisfying test accuracy and speed are achieved.…”