This thesis focuses on the development of an automated solution that can detect defects in chrome faucets by utilizing modern machine-learning techniques. The main goal was to optimize the accuracy and efficiency of detecting surface defects, therefore enhancing quality assurance in manufacturing operations.

The project involved collecting a dataset of chrome faucet images, which was subsequently enhanced using methods such as horizontal flips, rotations, brightness modifications, and the addition of Gaussian noise to enhance diversity. The augmented dataset was utilized to train CNN and SVM for defect detection.

A real-time defect detection solution has been developed with the ability to capture images from a camera and analyze them to find any defects. The models were assessed according to their accuracy and resilience in detecting surface defects under various circumstances. The results showed that the installed setup effectively decreases the need for manual inspections and offers consistent and accurate defect identification.

Ultimately, this work has effectively created a defect detection solution based on machine learning that can be seamlessly incorporated into factory settings to improve quality control. The results emphasize the possibility of utilizing CNN and SVM for automated defect detection, paving the way for more consistent and precise quality assurance procedures in industrial applications.