Article: Development of a key dimension inspection and error correction system for product design drawings combining computer vision and YOLOv10 Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.10 pp.46 - 59 Abstract: To address the challenges of large detection errors and difficult error correction in detecting key dimensions in paper product design drawings, we developed a novel object detection method based on YOLO v10. We added coordinate and attention mechanisms, along with a variable convolutional module, to the original method to optimise feature fusion. Furthermore, we optimised the loss function to better suit the specific application of product design drawings. Experimental results show that our proposed model has significant advantages over several classic object detection algorithms, achieving an average accuracy of 0.831 on the test set, significantly outperforming other benchmark models. We also conducted ablation studies on our model, revealing that the backbone network has the most significant impact on the overall model performance. Coordinate attention and variable convolution modules significantly boost performance. Our proposed model focuses on detecting and correcting key drawing dimensions, enhancing real-world applicability by advancing traditional models. Inderscience Publishers - linking academia, business and industry through research

Title: Development of a key dimension inspection and error correction system for product design drawings combining computer vision and YOLOv10

Authors: Liya Li; Fuguo Wu; Haijing Liu

Addresses: Department of Computer Engineering, Taiyuan Institute of Technology, Taiyuan, 030008, China ' Shanxi Aitplus Technology Co., Ltd., Taiyuan, 030006, China ' Department of Computer Engineering, Taiyuan Institute of Technology, Taiyuan, 030008, China

Abstract: To address the challenges of large detection errors and difficult error correction in detecting key dimensions in paper product design drawings, we developed a novel object detection method based on YOLO v10. We added coordinate and attention mechanisms, along with a variable convolutional module, to the original method to optimise feature fusion. Furthermore, we optimised the loss function to better suit the specific application of product design drawings. Experimental results show that our proposed model has significant advantages over several classic object detection algorithms, achieving an average accuracy of 0.831 on the test set, significantly outperforming other benchmark models. We also conducted ablation studies on our model, revealing that the backbone network has the most significant impact on the overall model performance. Coordinate attention and variable convolution modules significantly boost performance. Our proposed model focuses on detecting and correcting key drawing dimensions, enhancing real-world applicability by advancing traditional models.

Keywords: computer vision; YOLOv10; critical dimension detection; error correction; smart manufacturing.

DOI: 10.1504/IJRIS.2026.152543

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.10, pp.46 - 59

Received: 07 Nov 2025
Accepted: 05 Jan 2026
Published online: 26 Mar 2026 *

Title: Development of a key dimension inspection and error correction system for product design drawings combining computer vision and YOLOv10

Keep up-to-date