Version 1
: Received: 21 September 2024 / Approved: 23 September 2024 / Online: 23 September 2024 (11:58:05 CEST)
How to cite:
Gao, W.; Hu, H. Lightweight Neural Network Optimization for Rubber Ring Defect Detection. Preprints2024, 2024091770. https://doi.org/10.20944/preprints202409.1770.v1
Gao, W.; Hu, H. Lightweight Neural Network Optimization for Rubber Ring Defect Detection. Preprints 2024, 2024091770. https://doi.org/10.20944/preprints202409.1770.v1
Gao, W.; Hu, H. Lightweight Neural Network Optimization for Rubber Ring Defect Detection. Preprints2024, 2024091770. https://doi.org/10.20944/preprints202409.1770.v1
APA Style
Gao, W., & Hu, H. (2024). Lightweight Neural Network Optimization for Rubber Ring Defect Detection. Preprints. https://doi.org/10.20944/preprints202409.1770.v1
Chicago/Turabian Style
Gao, W. and Haijun Hu. 2024 "Lightweight Neural Network Optimization for Rubber Ring Defect Detection" Preprints. https://doi.org/10.20944/preprints202409.1770.v1
Abstract
Surface defect detection based on machine vision and convolutional neural networks (CNNs) is an important and necessary process that enables rubber ring manufacturers to improve production quality and efficiency. However, such automatic detection always consumes substantial computer resources to guarantee detection accuracy. To solve this problem, in this paper, we present a CNNs optimization algorithm based on the Ghost module. First, we replace the convolutional layer with the Ghost module in CNNs so that feature maps can be generated using cheaper linear operations. Second, we use an optimization method to obtain the best replacement of the Ghost module to achieve a balance between computer resource consumption and detection accuracy. Finally, we use an image preprocessing method that includes inverting colors. We integrated this algorithm into YOLOv5, which we trained on a dataset with 122 images of rubber ring surface defects. Compared with the original network, the network size decreased by 30.5% and the computational cost decreased by 23.1% whereas average precision only decreased by 1.8%. Additionally, the network's training time decreased by 16.1% as a result of preprocessing. These results show that the proposed approach greatly helped practical rubber ring surface defect detection.
Keywords
optimization; neural network; lightweight; rubber ring
Subject
Computer Science and Mathematics, Artificial Intelligence and Machine Learning
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
