Paper Title: YOLO v8 Based Anomaly Detection Model for Solar Photovoltaic Manufacturing Execution System
Conference Name: ICMIC 2024
Abstract: By exploring the efficacy of YOLO v8 in identifying
and classifying anomalies, this research contributes to advance-
ments in anomaly detection methodologies within the solar PV
sector. This paper proposed a YOLOv8-based anomaly detection
system tailored for pinpointing defects within solar PV manu-
facturing processes. The dataset comprises 921 surface images of
photovoltaic modules during factory manufacture, sourced from
the Solar Panel Fault Detection Dataset published by Solar Vision
Inc. comprising 5 categories of defects. We divided the dataset
into three sets: training,testing, and validation, with proportions
of 87%, 9%, and 5%, respectively. The model training involved
training the YOLOv8 model on the prepared dataset to identify
image features of various defects on the surface and accurately
locate them In terms of accuracy, sensitivity, and timeliness, the
result shows that our model achieved superior detection precision
of 100%, sensitivity of 99.9%, F1-score of 87.2% for solar PV
defects in a manufacturing execution system. The high detection
accuracy undoubtedly makes our model well suited for real-time
solar PV defects detection.