A Lightweight Cutting Tool Recognition Model Based on LResNet-ELA

View at publisher Summarise with VIRAT
Authors: Xinwei Wang, Hao Tang, Tianxu Wang, Lei Feng, Miao Wang
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Introduction: In intelligent industrial manufacturing, the rapid and accurate identification of cutting tools is crucial for production efficiency and processing quality. However, practical applications often face challenges such as the scarcity of dedicated tool datasets and the difficulty in balancing recognition accuracy with lightweight deployment on embedded devices. Methods: This paper proposes a lightweight cutting tool recognition model based on the LResNet- ELA architecture. The model integrates MobileNet's depthwise convolution to reduce parameters and computational costs, ResNet's residual structure to address deep network training issues, and the ELA mechanism to focus on key tool features. A dataset of 4207 cutting tool images was collected using industrial cameras, covering typical tool types, including end mills, face mills, taps, and drills. Results: The proposed LResNet-ELA model achieved a recognition accuracy of 99.48%, with a parameter count of only 0.23 million, ensuring both high performance and efficiency. These results indicate that the architecture is well-suited for real-time deployment in embedded devices. Discussion: The results demonstrate that the LResNet-ELA model offers an effective solution for cutting tool identification, balancing high recognition accuracy with resource-efficient deployment. However, the model's performance could be further validated under different industrial environments and tool types to ensure generalizability. Conclusion: This lightweight architecture provides a feasible solution for real-time tool identification on embedded devices in machine tool environments, offering a balance between high recognition accuracy and resource-efficient deployment.

Fields of Study & Subjects

Advanced Neural Network Applications Industrial Vision Systems and Defect Detection Engineering Technology and Methodologies

Publication details

Journal / source
Recent Patents on Mechanical Engineering
Volume
19
Publication date
30 March 2026
Language
EN