Intelligent Defect Detection of Design Materials Based on Edge AI
DOI:
https://doi.org/10.67541/jdf2609Keywords:
Edge AI; Defect detection; Lightweight network; Multi-scale attention fusion; Adaptive inferenceAbstract
Aiming at the problems such as the imbalance between accuracy and efficiency, insufficient adaptability of multi-scale defects and limited inference resources in the deployment of design material defect detection at the edge end, this paper proposes a lightweight detection method based on edge AI. In the training stage, multi-branch topology is used to enhance feature expression, and in the inference stage, multi-branch convolution is merged into a single standard convolution by algebraic fusion. Under the condition that the number of parameters is only 2.2M and the amount of computation is 2.6 GFLOPs, the detection accuracy is significantly improved. The bidirectional feature pyramid and lightweight channel-spatial joint attention fusion module are designed to effectively cope with the challenges of defect scale diversity, low contrast and inter-class similarity, and the recall rate of small-scale defect reaches 82.4%. An early departure inference mechanism based on input complexity and a dynamic computing path guided by confidence are constructed, and combined with quantization-aware training and operator fusion optimization, 18ms end-to-end inference latency and 55.6 FPS frame rate are achieved on Jetson Xavier NX, with power consumption of only 2.4 W. Experiments on NEU-DET, DAGM2007 and self-built industrial datasets show that the proposed method mAP@0.5 reaches 75.2%, which outperforms the mainstream lightweight detection network with more than three times the number of parameters, and shows superior performance in a variety of defect types and scales. It provides a feasible technical path for real-time, safe and low-cost detection in intelligent manufacturing scenarios.
References
[1] Hassan, N. M., Hamdan, A., Shahin, F., Abdelmaksoud, R., & Bitar, T. (2023). An artificial intelligent manufacturing process for high-quality low-cost production. International Journal of Quality & Reliability Management, 40(7), 1777-1794. https://doi.org/10.1108/IJQRM-07-2022-0204
[2] Zhao, T., Chen, G., Suraphee, S., Phoophiwfa, T., & Busababodhin, P. (2025). A hybrid TCN-XGBoost model for agricultural product market price forecasting. PLoS One, 20(5), e0322496. https://doi.org/10.1371/journal.pone.0322496
[3] Zhang, D., Zhou, S., Zheng, Y., & Xu, X. (2025). Review on application of machine vision-based intelligent algorithms in gear defect detection. Processes, 13(10), 3370. https://doi.org/10.3390/pr13103370
[4] Wang, X., Tang, Z., Guo, J., Meng, T., Wang, C., Wang, T., & Jia, W. (2025). Empowering edge intelligence: A comprehensive survey on on-device AI models. ACM Computing Surveys, 57(9), 1-39. https://doi.org/10.48550/arXiv.2503.06027
[5] Zhao, T., Chen, G., Pang, C., & Busababodhin, P. (2025). Application and performance optimization of SLHS-TCN-XGBoost model in power demand forecasting. Comput. Model. Eng. Sci, 143(3), 2883-2917. http://dx.doi.org/10.32604/cmes.2025.066442
[6] Shuvo, M. M. H., Islam, S. K., Cheng, J., & Morshed, B. I. (2022). Efficient acceleration of deep learning inference on resource-constrained edge devices: A review. Proceedings of the IEEE, 111(1), 42-91. https://doi.org/10.1109/JPROC.2022.3226481
[7] Wang, Z., Zhang, Z., Su, T., Ding, Z., & Zhao, T. (2024, December). Research on supply chain network optimisation based on the CNNs-BiLSTM model. In 2024 International Conference on Information Technology, Comunication Ecosystem and Management (ITCEM) (pp. 197-202). IEEE. https://doi.org/10.1109/ITCEM65710.2024.00044
[8] Abubakar, M., Che, Y., Zafar, A., Al-Khasawneh, M. A., & Bhutta, M. S. (2025). Optimization of solar and wind power plants production through a parallel fusion approach with modified hybrid machine and deep learning models. Intelligent Data Analysis, 29(3), 808-830. https://doi.org/10.1177/1088467x241312592
[9] Zhao, T., Chen, G., Pang, C., Li, L., & Busababodhin, P. (2026). Forecasting global agricultural trade imbalances using a hybrid deep learning and gradient boosting framework. Discover Computing, 29(1), 443. https://doi.org/10.1007/s10791-026-10367-8
[10] Dalal, S., Lilhore, U. K., Simaiya, S., Radulescu, M., & Belascu, L. (2024). Improving efficiency and sustainability via supply chain optimization through CNNs and BiLSTM. Technological Forecasting and Social Change, 209, 123841. https://doi.org/10.1016/j.techfore.2024.123841
[11] Liu, Y., Gong, M., Chen, G., & Zhao, T. (2025, July). CNN-LSTM-based spatio-temporal prediction model for photovoltaic power generation. In Proceedings of the 2025 2nd International Conference on Big Data and Digital Management (pp. 570-575). https://doi.org/10.1145/3768801.3768892
[12] Patil, S. P., Mallika, R. M., Sekhar, A. C., Josephson, P. J., & Jacob, T. P. (2026, May). Hybrid CNN-BiLSTM Model for the Enhancement of Social Sustainability in Manufacturing Supply Chains. In 2026 International Conference on Intelligent and Sustainable Electronics & Computing Technologies (INSECT) (pp. 1-6). IEEE. https://doi.org/10.1109/insect68872.2026.11663829
[13] Peng, M., Hu, J., & Zhao, T. (2026, March). Hybrid ARIMA-LSTM model for international trade logistics demand forecasting. In IET Conference Proceedings CP982 (Vol. 2026, No. 9, pp. 243-247). Stevenage, UK: The Institution of Engineering and Technology. https://doi.org/10.1049/icp.2026.2978
[14] Wang, Y., & Liang, X. (2025). Application of reinforcement learning methods combining graph neural networks and self-attention mechanisms in supply chain route optimization. Sensors, 25(3), 955. https://doi.org/10.3390/s25030955
[15] Zhao, T., Chen, G., Pang, C., Seenoi, P., Papukdee, N., Busababodhin, P., & Du, Y. (2025). Hybrid convolutional neural network-graph attention network-gradient boosting decision tree model for seismic impedance inversion prediction. J Seismic Explor, 34(5), 81-98. https://doi.org/10.36922/JSE025310051
[16] Zhao, T., Chen, G., & Busababodhin, P. (2026). An intelligent expert system for logistics disruption prediction and mitigation in global supply chains: Integrating graph-based risk inference with ensemble forecasting. Scientific Reports. https://doi.org/10.1038/s41598-026-61617-0
[17] Qu, J., Qian, Z., & Pei, Y. (2021). Day-ahead hourly photovoltaic power forecasting using attention-based CNN-LSTM neural network embedded with multiple relevant and target variables prediction pattern. Energy, 232, 120996. https://doi.org/10.1016/j.energy.2021.120996
[18] Chen, G., Zhao, T., Pang, C., & Busababodhin, P. (2026). Integrated CNN–LSTM–XGBoost hybrid model predicts shale oil seismic attributes and global oil price trends. Scientific Reports, 16, 25588. https://doi.org/10.1038/s41598-026-55910-1
[19] Ma, J., Huo, M., Han, J., Liu, Y., Lu, S., & Yu, X. (2025). Integrated CNN‐LSTM for photovoltaic power prediction based on spatio‐temporal feature fusion. Engineering Reports, 7(1), e13088. https://doi.org/10.1002/eng2.13088
[20] Zhao, T., Chen, G., Pang, C., Seenoi, P., Papukdee, N., & Busababodhin, P. (2025). Time-lapse earthquake difference prediction based on physics-informed long short-term memory coupled with interpretability boosting. Journal of Seismic Exploration, 34(3), 25. https://doi.org/10.36922/JSE025310049
[21] Zhang, X., Weng, Z., Zhu, P., Han, X., Zhu, J., & Jiao, L. (2024). ESDINet: Efficient shallow-deep interaction network for semantic segmentation of high-resolution aerial images. IEEE Transactions on Geoscience and Remote Sensing, 62, 1-15. https://doi.org/10.1109/tgrs.2024.3351437
[22] Zhao, Y., Liu, Q., Su, H., Zhang, J., Ma, H., Zou, W., & Liu, S. (2024). Attention-based multiscale feature fusion for efficient surface defect detection. IEEE Transactions on Instrumentation and Measurement, 73, 1-10. https://doi.org/10.1109/tim.2024.3372229
[23] Su, J., Luo, Q., Yang, C., Gui, W., Silvén, O., & Liu, L. (2024). PMSA-DyTr: Prior-modulated and semantic-aligned dynamic transformer for strip steel defect detection. IEEE Transactions on Industrial Informatics, 20(4), 6684-6695. https://doi.org/10.1109/tii.2023.3347747
[24] Hamed, O., Bakkali, S., Blaschko, M., Moens, S., & Van Landeghem, J. (2024, August). Multimodal adaptive inference for document image classification with anytime early exiting. In International Conference on Document Analysis and Recognition (pp. 270-286). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-70546-5_16
[25] Zhong, H., Fu, D., Xiao, L., Zhao, F., Liu, J., Hu, Y., & Wu, B. (2024). STFE-Net: A multi-stage approach to enhance statistical texture feature for defect detection on metal surfaces. Advanced engineering informatics, 61, 102437. https://doi.org/10.1016/j.aei.2024.102437
[26] Xiang, Z., Jia, J., Zhou, K., Qian, M., & Wu, W. (2025). Block-wise feature fusion for high-precision industrial surface defect detection. The Visual Computer, 41(11), 9277-9295. https://doi.org/10.1007/s00371-025-03927-4
[27] Cai, X., Wang, Y., & Zhang, L. (2022). Optimus: An operator fusion framework for deep neural networks. ACM Transactions on Embedded Computing Systems, 22(1), 1-26. https://doi.org/10.1145/3520142
[28] Li, Y., Ren, Y., Wang, M., Zhang, Y., Zhang, X., & Deng, J. (2026). Deployment method of deep object detection models integrating model pruning and data reuse for reconfigurable architectures: Y. Li et al. The Journal of Supercomputing, 82(12), 606. https://doi.org/10.1007/s11227-026-08749-2
[29] Yao, G., Zhu, S., Zhang, L., & Qi, M. (2024). HP-YOLOv8: high-precision small object detection algorithm for remote sensing images. Sensors, 24(15), 4858. https://doi.org/10.3390/s24154858
[30] Wang, C., Zheng, B., & Li, C. (2025). Efficient traffic sign recognition using YOLO for intelligent transport systems. Scientific reports, 15(1), 13657. https://doi.org/10.1038/s41598-025-98111-y
Published
Data Availability Statement
The data that support the findings of this study are available upon request from the corresponding authors, L.M.
Funding information
This research was financially supported by the Key Project of the Hubei Provincial Department of Education for Philosophy and Social Science Research, “Research on Digital Innovation Design from a Metaverse Perspective” (No. 23D026).
Issue
Section
License
Copyright (c) 2026 Jiayu Du, Mengyu Liu (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
