Federation Learning for Vision Based Product Quality Inspection
Nischitha T S1, Harshitha N2, Harshitha N G3,Manoj T4 ,Rohith D M5
1Nischitha T S Information Science and Engineering & RR Institute of Technology
2Harshitha N Information Science and Engineering & RR Institute of Technology
3Harshitha N G Information Science and Engineering & RR Institute of Technology
4Manoj T Information Science and Engineering & RR Institute of Technology
5Rohith D M Information Science and Engineering & RR Institute of Technology
Abstract - In modern manufacturing industries, maintaining consistent product quality is a major challenge. Vision-based product quality inspection systems help in identifying defects by analyzing images of products. However, traditional machine learning approaches require collecting data from all production units at one place, which can raise concerns related to data privacy, security, and high communication cost. To overcome these limitations, this project adopts a federated learning approach for vision-based product quality inspection. In the proposed system, image data is processed locally at different inspection units, and only the learned model parameters are shared with a central server instead of raw images. This ensures data privacy while allowing the system to learn from multiple sources. The trained federated model is capable of detecting defects such as surface irregularities, shape deviations, and visual inconsistencies with improved accuracy. By combining computer vision techniques with federated learning, the system achieves efficient defect detection while reducing data transfer and protecting sensitive manufacturing information. This approach proves to be a reliable, scalable, and secure solution for automated product quality inspection in industrial environments. The integration of AI-driven inspection with IoT-based actuation establishes a closed-loop quality control system that significantly minimizes human error, reduces inspection time, and enhances overall manufacturing throughput. By employing techniques such as image preprocessing, feature extraction, and model optimization, the system achieves high detection accuracy even in varying lighting and background conditions. The proposed solution not only enhances consistency and reliability in quality assurance but also enables scalability across diverse product lines with minimal hardware or algorithmic adjustments. Ultimately, this project demonstrates how the fusion of machine learning and IoT can revolutionize industrial inspection processes, transforming traditional manual quality checks into intelligent, automated, and data-driven operations. The results affirm that the ML-IoT integrated inspection system substantially improves productivity, consistency, and precision in Wheels assembly operations.
Key Words: Fault detection in the product, Federation Learning, Vision Based Inspection System, Voice Feedback.
