Name
Yijia Xue
Major
Data Science
Class
2023
About
| This E-portfolio provides all the products related to the Signature Work. |
| This E-portfolio provides all the products related to the Signature Work. |
| This project presents a novel approach to industrial visual defect detection and localiza‐ tion using an interpretable deep learning‐based algorithm, Interpretable Aware Au‐ toencoder (IAAE). The proposed algorithm aims to overcome the limitations of tra‐ ditional anomaly detection methods by leveraging the power of deep learning while maintaining interpretability and scalability. The IAAE algorithm is designed to detect and localize visual defects in various industrial products, including textiles, electronic components, pharmaceuticals, and construction materials. The project provides a detailed description of the IAAE algorithm, including its architec‐ ture, training procedure, and evaluation metrics. The experimental findings obtained through the proposed network are presented in this paper. The datasets used for train‐ ing and testing the IAAE algorithm are described in detail. The comparative results show that the proposed algorithm outperforms baseline anomaly detection methods in terms of accuracy and interpretability. The qualitative results obtained from the experiments demonstrate that the IAAE al‐ gorithm can effectively detect and localize visual defects in various industrial prod‐ ucts. The proposed algorithm has significant potential for various industrial applica‐ tions, including quality control, product inspection, and defect prevention. Overall, this paper provides a comprehensive analysis of the proposed approach to industrial visual defect detection and localization using an interpretable deep learning‐based al‐ gorithm |
昆山杜克大学 苏ICP备16021093号