Advances in Sensor Research: University of Valencia Researchers Evaluate Machine Learning and Deep Learning Architectures

Strategic Insights -

Research published by the University of Valencia has led to significant advancements in sensor research, particularly in the area of industrial surface defect detection. The study, conducted by Azeddine Mjahad and Alfredo Rosado-Munoz, employed classical Machine Learning (ML) algorithms and deep learning architectures to evaluate the effectiveness of statistical parameters and network architectures in detecting surface defects. The findings indicate that a combination of statistical descriptors with ML and DL architectures provides a robust and scalable solution for automated, non-destructive surface defect detection.

Key Takeaways:

  • The study evaluated classical ML algorithms using extracted statistical parameters and deep learning architectures, including ResNet50, Capsule Networks, and a 3D Convolutional Neural Network (CNN3D).
  • The Dataset Original was used to train ML models, achieving high performance: RF reached 99.4 ± 0.2% precision and 99.4 ± 0.2% sensitivity, GB 96.0 ± 0.2% precision and 96.0 ± 0.2% sensitivity.
  • ResNet50 trained with extracted parameters reached 98.0 ± 1.5% accuracy and 98.2 ± 1.7% F1-score.
  • Capsule-based architectures achieved the best results, with ConvCapsuleLayer reaching 98.7 ± 0.2% accuracy and 100.0 ± 0.0% precision for the normal class, and 98.9 ± 0.2% F1-score for the affected class.
  • CNN3D applied on 3D image inputs reached 88.61 ± 1.01% accuracy and 90.14 ± 0.95% F1-score.
  • The research demonstrated consistent high performance using the Dataset Expanded with ML and PCA-selected features, achieving 99.4 ± 0.2% precision and 99.4 ± 0.2% sensitivity with Random Forest, K-Nearest Neighbors, and SVM.

Statistics:

  • The average execution time per image was as low as 3.69 x 10^-4 s/image, supporting potential real-time industrial application.
  • The study evaluated 99.4 ± 0.2% precision and 99.4 ± 0.2% sensitivity for Random Forest and K-Nearest Neighbors using the Dataset Expanded.

Sources:

  • Robust Industrial Surface Defect Detection Using Statistical Feature Extraction and Capsule Network Architectures. Sensors, 2025,25(19):6063.
  • Our news journalists report that more information may be obtained by contacting Azeddine Mjahad, GDDP, Department Electronic Engineering, School of Engineering, University of Valencia, 46100 Burjassot, Valencia, Spain.