A Hard Voting Policy-Driven Deep Learning Architectural Ensemble Strategy for Industrial Products Defect Recognition...

Area of Science:

• Computer Vision
• Machine Learning
• Industrial Automation

Background:

• Traditional industrial inspection methods are complex, time-consuming, and error-prone, hindering effective quality control.
• Existing intelligent models often sacrifice real-time performance for accuracy, limiting their practical application in manufacturing.

Purpose of the Study:

• To develop an efficient, rapid, and intelligent model for industrial product fault recognition and classification.
• To enhance visual inspection and quality control processes through improved defect detection.

Main Methods:

• Proposed an ensemble deep learning framework utilizing a model architectural voting policy.
• The voting policy considers model optimality, efficiency, and performance accuracy for feature learning.
• Validated the framework on three publicly available industrial product datasets.

Main Results:

• Demonstrated a significant increase in fault recognition and classification performance for industrial products.
• The proposed model achieved remarkable results in identifying and categorizing defects.
• The ensemble approach effectively computed and learned hierarchical features in industrial artifacts.

Conclusions:

• The developed deep learning framework offers an efficient and accurate solution for industrial defect recognition.
• The model's voting policy enhances performance, making it suitable for real-time quality control applications.
• This approach significantly improves industrial product inspection and overall manufacturing quality.