ABSTRACT:

 

The MVTec AD (Anomaly Detection) dataset is a comprehensive, real-world dataset designed for benchmarking unsupervised anomaly detection algorithms, specifically in industrial and manufacturing environments. This research focuses on applying state-of-the-art YOLOv8 (You Only Look Once version 8) for anomaly detection, an advanced object detection model that excels in identifying defects and irregularities in images. The dataset includes images with a variety of anomalies, such as scratches, deformations, and other manufacturing defects typically found in industrial products like electronics, machinery, and textiles. The objective of this work is to develop an automated anomaly detection system that can accurately identify these defects using YOLOv8, offering a robust solution for quality control in industrial settings.The front-end is built using Streamlit, an easy-to-use framework that enables the creation of interactive and visually appealing web applications for anomaly detection. YOLOv8 is utilized to detect and classify anomalies in real-time, producing efficient results for large datasets. This integrated approach showcases how advanced deep learning algorithms can be employed with scalable tools to improve the efficiency and accuracy of anomaly detection in industrial applications, reducing manual inspection efforts and improving product quality.

 

Keywords—Anomaly Detection, YOLOv8, MVTec AD, Deep Learning, Quality Control